- Email: [email protected]
The effect of feeding and management practices on calving rate in dairy herds
Animal Reproduction Science 74 (2002) 133–150
The effect of feeding and management practices on calving rate in dairy herds J. Fahey a,1 , K. O’Sullivan b , J. Crilly a , J.F. Mee a,∗ a
b
Teagasc, Dairy Production Department, Dairy Production Research Centre, Moorepark, Fermoy, Co. Cork, Ireland Statistical Laboratory, Department of Statistics, University College Cork, Cork, Ireland
Received 29 November 2001; received in revised form 15 May 2002; accepted 23 August 2002
Abstract The objective of this study was to examine the effects of nutrition and management practices on reproductive performance in 31 Irish dairy herds participating in the Moorepark Dairy Management Information System (DairyMIS) during the period 1991–1998. Fifty variables relating to herd reproductive indices, calving events, stocking rate, disease, concentrate feeding, fertiliser usage, milk production and economic performance were studied using factor analysis. A factor analysis, followed by varimax rotation, identified 13 factors, which accounted for 83% of the total variance of the original variables. A regression model was used to predict calving rate from the orthogonal factor scores identified by factor analysis. Calving rate was defined as the proportion of services, for which an outcome was known, which resulted in a subsequent calving. Year, farm code and factor 3 (labelled herd size) together accounted for 40% of the observed variation in calving rate. The factor scores for factor 3 (herd size) were plotted against calving rate and because the plot was not linear, it was decided that dividing the factor scores into quartiles would produce a better fitting model. The factor scores for herd size were sorted and assigned to four equal categories (n = 47 per category), from lowest to highest. The ranges in herd size according to category were as follows: category 1 (34–96), category 2 (47–103), category 3 (66–152) and category 4 (108–359). The calving rate (%), (±S.E.) was 67.0 ± 2.5 for category 1, 61.8 ± 1.8 for category 2, 56.9 ± 1.5 for category 3 and 53.2 ± 2.85 for category 4. Using pair-wise comparisons, calving rates differed significantly (P < 0.05) between all categories except between categories 3 and 4 (P > 0.05). Herd-level milk production was not associated with calving rate indicating that good management
∗ Corresponding author. Tel.: +353-25-42387; fax: +353-25-42340. E-mail address: [email protected] (J.F. Mee). 1 Present address: Department of Veterinary Science, Victorian Institute of Animal Science, University of Melbourne, 600 Sneydes Road, Werribee, Victoria 3030, Australia.
0378-4320/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 4 3 2 0 ( 0 2 ) 0 0 1 9 2 - 6
134
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
may overcome any adverse effects of high milk production on reproductive performance. Larger herds, in combination with other associated herd characteristics, are likely to have poorer calving rates than medium or small herds. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Factor analysis; Fertility; Herd size; Nutrition; Management; Dairy cows
1. Introduction Optimising reproductive efficiency is essential in maintaining profitability in a dairy herd. The economic advantages of a compact calving pattern in seasonal milk production, which allows 80% of milk to be produced from spring grass, have been quantified by Dillon (1996). While genetic improvement of dairy cows has led to a dramatic increase in milk yield (Funk, 1993), genetic correlation estimates indicate an antagonistic relationship between milk production and fertility (Van Arendonk et al., 1991; Hoekstra et al., 1994; Pryce et al., 1998). Reproductive performance of Irish dairy herds has been assessed previously using research databases (O’Farrell and Crilly, 1998; Mee et al., 1999; Buckley et al., 2000) and surveys (Crowley et al., 1967; White and O’Farrell, 1972; Roche et al., 1978, Cunningham et al., 1978; Cunningham and O’Byrne, 1980). In the studies undertaken between 1967 and 1980, calving rate to first service was reported to be 60–69%. In the study reported by Mee et al. (1999), calving rate to first service fell significantly (0.9% per year) from 59% in 1991 to 54% in 1998. The potential effects of nutritional and management factors and their changes over time on reproductive performance in spring-calving Irish dairy herds have not been examined previously. Evaluation of management across a large number of herds can be difficult. Data may be incomplete and not comparable. A dairy herd is a complex and dynamic system in which input and output are related to management strategies and animal status in a complex manner (Eneveloldsen et al., 1995). Many management indicators are strongly interrelated making it difficult to specify explicitly which variables are dependent and which are independent in a traditional analysis (Eneveloldsen et al., 1996). Data reduction techniques, such as principal component analysis and factor analysis, have been successfully used to evaluate underlying relationships among multiple traits or management types measured simultaneously (Sieber et al., 1987; Holloway et al., 1990; Price and Wallach, 1991; Hurnik et al., 1994; Eneveloldsen et al., 1996; Shahin et al., 1998). The objective of this study was to establish the association between dairy herd management indicators (calving events, stocking rate, milk yield and milk composition, concentrate feeding, fertiliser usage, disease and economic performance) and calving rate in herds representative of seasonal milk production systems and participating in the Moorepark Dairy Management Information System (DairyMIS) during the period 1991– 1998.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
135
2. Materials and methods 2.1. Data Information on farm management practices was compiled from the DairyMIS database as described by Crosse (1985). The database consisted of 32 farms (herds) recorded over an 8-year period from 1991 to 1998 (284 herd years) although some of the farms were not present for the entire 8-year period. Calving rate (CR) was defined as the proportion of services, for which an outcome was known, which resulted in a subsequent calving. Where a service was followed by a subsequent calving at an interval of no more than 300 days, the animal was deemed to have conceived to that service. Where a service was followed by a further service event at an interval of at least 5 days, the initial service was considered to have failed. Where two services were recorded within an interval of 4 days the second service was ignored and any conception was attributed to the first event. If a calving event occurred more than 300 days after the last recorded service event, the cow was assumed to have conceived to an unrecorded service (O’Farrell and Crilly, 1998). If greater than 10% of cows in a given year calved to an unrecorded service the records were considered to be unreliable and the data for that herd-year were excluded from the analysis. Where more than 30% of cows served in a given year failed to calve in the following year the data for that herd-year were again excluded from the analysis. Very low calving rates tended to be associated with events such as partial depopulations following outbreaks of notifiable diseases and their inclusion in the database could be a major source of bias. When herd-years were excluded as described, the dataset consisted of 31 herds and 188 herd-years. The participating farms were commercial dairy herds with an average of 108 cows (range 34–359) producing an average milk yield of 5000 kg per cow per year (range 4000–7000 kg). The herds were predominantly spring/summer-calving with more than three-quarters of the herds having ≥85% of their cows calving between January and June. The farms involved in the study were situated in counties Cork, Waterford and Tipperary in southern Ireland. 2.2. DairyMIS variables The variables (all continuous) recorded in the DairyMIS database relate to herd reproductive indices, calvings, stocking rate, disease, concentrate feeding, fertiliser usage, milk production and economic indices (Tables 1–8). Herd-year was used as the unit of analysis. The name, abbreviation (e.g. stocking rate per hectare (name), SRHA (abbreviation)) and definition for the 50 variables in the database are presented. 2.3. Statistical analyses 2.3.1. Factor analysis A common factor analysis model using the Proc Factor procedure of SAS (1989) computer software was used to condense the information contained in the 50 original variables into a smaller set of dimensions (factors). The correlation matrix of each variable with all other variables was the basis for factor extraction. The variance accounted for by each factor,
136
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 1 Variables relating to herd reproductive indices recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Calving to first service interval (CSE)
The mean interval in days from last calving date to first service date. The proportion of artificial inseminations carried out by a commercial technician. It was possible to categorise inseminations as commercial, DIY or natural service based on information supplied by the farmer, together with inseminator codes. The date of the first service event in days of year (1–365). If no further cows were served within 5 days of the first service event then the first service event was ignored and the next service date was taken as the beginning of the breeding season provided two further service events occurred within a 10-day period. In order to overcome the calendar year-on-year effect (i.e. 31 December = day 365 and 1 January = day 1), when beginning of the breeding season day was >200 it was taken as that day of year minus 365. This gave a continuous breeding season from July to July.
Commercial AI (COM)
Beginning of breeding season (Bosday)
its eigenvalue, was used to determine its significance and whether it would be retained for further analysis (Hair et al., 1998). Because the original factor loadings (the correlations between the underlying factors and the original variables) were not readily interpretable, an orthogonal transformation matrix Table 2 Variables relating to cows calving recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total number of calvings (NOCA) Herd size (HS) Cows calving percentage (COCAPC) Mean calving date (DAYC)
Total number of cows calving Total number of cows in the herd Number of calvings/cows in herd × 100 If calving day was greater than day 200 then mean calving day was set to calving day minus 365. Similar to beginning of breeding season day this was done to take into account the difference between 31 December (day 365) and 1 January (day 1) where cows calve from one year to the next. Number of cows calving from January to June/total number of calvings Number of cows calving from January to March/total number of calvings Number of cows calving from April to June/total number of calvings Number of cows calving from July to September/total number of calvings Number of cows calving from October to December/total number of calvings Cows in milk/cows in herd × 100
Cows calving from January to June (CA16) Cows calving from January to March (CA13) Cows calving from April to June (CA46) Cows calving from July to September (CA79) Cows calving from October to December (CA1012) Cows in milk percentage, lactation length (INMILK)
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
137
Table 3 Variables relating to stocking rate recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Culling/replacement rate (CULLPC) Number of hectares (HA) Stocking rate/HA (SRHA)
Number of cows sold/cows in herd × 100 Number of hectares allocated to the dairy enterprise Area of land devoted to dairy enterprise (HA)/total livestock units (number of cows multiplied by 1, plus heifers in calf multiplied by 0.7, plus cattle under 1 year multiplied by 0.3, plus cattle between 1 and 2 years multiplied by 0.7, plus cattle over 2 years multiplied by 1, plus cull cows multiplied by 1, plus any other livestock units (horses, sheep, deer) Cows in herd/total livestock units × 100
Cows as percentage of total livestock units (COTOTLU) Lactation number (LN) Silage cut (Cut)
Average number of lactations for cows in herd Percentage of farm cut for first + second + third cut grass silage
Table 4 Variables relating to disease recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Calf mortality (CARIP)
Number of births dead/births total × 100. Number of births dead includes certain abortion, uncertain abortion, stillbirth, dead in 24 h, dead in 48 h and dead between 3 days and 3 weeks Taken from milk test (weighted average over time) Taken from milk test (weighted average over time)
Somatic cell count (SCC) Total bacteria count (TBC)
Table 5 Variables relating to concentrate feeding recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Concentrate fed per cow per year (CPC) Residual concentrate feeding per year (RMEAL)
Total tonnes of concentrates used/cows in herd Residual concentrate feeding when milk yield was taken into account The residuals are calculated from the model: CPC = constant + yield. Concentrate fed per cow/(cows in herd−cows in milk) Concentrate fed per cow/cows in milk Concentrate fed per cow/litres per cow Utilised metabolisable energy (UME)/HA. UME was calculated assuming that 25,000 MJ of energy was required per cow per year for maintenance, 5 MJ of energy was required to produce a litre of milk, and concentrates provide 11.5 MJ/kg UME/cows in herd
Concentrate fed/cows dry per year (MPCD) Concentrate fed/cows in milk per year (MPCIM) Concentrate per litre per year (MPL) Utilised metabolisable energy/HA per year (UHA)
Utilisable metabolisable energy per cow per year (UCOW)
138
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 6 Fertiliser variables recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total units of nitrogen/HA per year (NHA) Total units of nitrogen/HA from March to May (NHA35)
Units of nitrogen fertiliser used/HA Units of nitrogen fertiliser used from March to May/HA Residual units of nitrogen fertiliser/HA. The residuals are calculated from the model: NHA = constant + SRHA + Cut. Units of phosphorus fertiliser/HA Units of potassium fertiliser/HA Units of sulphur fertiliser/HA
Residual units of nitrogen/HA per year (RNHA)
Total units of phosphorus fertiliser/HA per year (PHA) Total units of potassium fertiliser/HA per year (KHA) Total units of sulphur fertiliser/HA per year (SULHA)
Table 7 Milk production variables recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total milk produced per year (TOTMK) Milk fat yield in litres per year (YF) Milk protein yield in litres per year (YP) Milk fat percentage per year (FAT) Milk protein percentage per year (PROT) Milk production per cow in litres per year (Yield) Yield of milk/HA per year (Yieldha) Yield of milk between April and June (Yield46)
Total herd milk production Total milk fat production in litres Total milk protein production in litres Milk fat percentage Milk protein percentage Total milk produced/cows in herd Milk production per cow in litres × stocking rate/HA Total milk produced between April and June/cows in milk between April and June
Table 8 Economic indices recorded in DairyMIS including variable name, abbreviation and definition Variable
Definition
Concentrate cost per year (CCSTAND) Milk price (MPSTAND)
Concentrate cost is equal to the total tonnes used × cost per tonne. Milk price standardised within year (Concentrate cost per tonne × total concentrates fed)/cows in herd + (other feed costs per cow) (Concentrate cost per tonne × total concentrates fed)/cows in herd + (other feed costs per cow)/milk yield per cow (Milk value−concentrate cost−other feed costs−fertiliser cost)/cows in herd (Milk value−concentrate cost−other feed costs−fertiliser cost)/HA Milk value−concentrate cost−other feed costs−fertiliser cost/yield
Feed costs per cow per year (FCCSTAND)
Feed costs per litre per year (FCLSTAND)
Margin over feed and fertiliser per cow per year (MOFPCSTA) Margin over feed and fertiliser/HA per year (MOFHASTA) Margin over feed and fertiliser per litre per year (MOFPLSTA)
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
139
140
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
141
was used to rotate the factor loadings (Hair et al., 1998). Rotation using an orthogonal transformation matrix (e.g. varimax rotation) maintains the independence of the factors (Gilbert and Bailey, 1991). After extraction of the initial factor solution (common factors), the matrix was rotated, using the varimax method (Kaiser normalisation), in order to obtain a clear pattern of loadings, thereby making the results more interpretable (Table 9). Each variable then had a high loading on a single factor and a small to moderate loading on other factors (Johnson and Wichern, 1982). When assessing factors, it is important that they make biological sense. Factors may be rotated by several techniques to simplify the factor structure and facilitate interpretation. If more than one common factor is identified in a factor analysis and an oblique rotation has produced correlations between the factors, the interrelations between these first-order factors can be described by a second factor analysis based on factor scores or interfactor correlations from the first-order analysis (Eneveloldsen et al., 1996). The second-order factor analysis then identifies a second set of common factors that explains the variation in the obliquely rotated first-order factors. As the study reported by Eneveloldsen et al. (1996) was similar in its data and objectives, second-order factor analysis was attempted in the current study. Four second-order factors were produced. However, these factors were uninterpretable, therefore, first-order factor analysis using orthogonal rotation was preferred. In the present study, the factors were labelled according to the pattern of factor loadings. Significant loadings for each factor were examined and, in general, the larger the absolute size of the factor loading for a variable, the more important the variable was in interpreting the factor. All the variables loading on a factor were considered including the size and sign of the loading. A determination was then made as to what the underlying factor represented (ACITS, 1995). 2.4. Regression analysis (factors) The scores of the 13 factors were used as determinants in a general linear model in order to identify which, if any, of the factors relating to feeding and farm management practices were associated with calving rate. Calving rate to all services was taken as the variable, which best represented herd reproductive performance. Calving rate was used as the dependent variable in a linear regression model (original model) which contained year, farm code and the scores for each of the 13 factors as determinants. Farm code was classified as a random effect and year as a fixed effect. The regression procedure utilised was backward elimination. Initially all factor scores were included in the model and the least significant factor scores were removed one at a time until each factor score in the model had an overall probability of <0.05. Once eliminated, factor scores were not allowed to be included again in the model building process. Goodness of fit was evaluated from residual plots and predictive ability was assessed by R2 values. Residual diagnostics did not indicate any concern for departures from the statistical assumptions of constant variability and normality. 3. Results The mean calving rate to all services across all herd-years was 58% while the lowest and highest herd-year calving rates ranged from 35 to 85%. The range in calving rate between
142
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 10 Descriptive statistics of calving rate to all services, 1991–1998 Calving rate to all services
1991
1992
1993
1994
1995
1996
1997
1998
Number of herds Mean Standard error Median Minimum Maximum
18 0.59 0.03 0.60 0.35 0.85
25 0.60 0.02 0.59 0.42 0.81
26 0.58 0.02 0.57 0.36 0.74
27 0.56 0.01 0.55 0.43 0.72
24 0.60 0.02 0.62 0.42 0.76
20 0.59 0.02 0.58 0.46 0.70
24 0.55 0.02 0.55 0.41 0.72
24 0.57 0.02 0.58 0.41 0.73
herd averages across years was 49–74%. The mean range in calving rate to all services within herd between years was 18% where the lowest was 0% and the highest 38%. Descriptive statistics of calving rate across herds between years are presented (Table 10). 3.1. Factor analysis Thirteen factors, or latent variables of the initial factor solution, accounted for 83% of the total variance in the 50 variables. The rotated factor loading ‘r’ is similar in its interpretation to the correlation coefficient. Factor loadings equivalent to correlation coefficients of 0.30 and greater are highlighted to indicate the main attributes of the different factors (Table 9). The data structure between the 50 original variables and the 13 factors is presented in Fig. 1. 3.1.1. Factor 1 (winter-calving/concentrate feeding) Factor 1 was the most clearly defined, with 16 of the 50 original variables loading significantly. Farms with a high score for factor 1 had a high proportion of cows calving in the winter (CA1012, r = 0.60) and a low proportion calving in early spring (CA13, r = −0.63; CA16, r = −0.64). As a consequence, mean calving date (DAYC, r = −0.58) and start of the breeding season (BOSDAY, r = −0.65) were early. Concentrate feeding (MPL, r = 0.97; C50PC, r = 0.95; RMEAL, r = 0.95; MPCIM, r = 0.92) was high which resulted in high feed costs (FCLSTAND, r = 0.90; FCCSTAND, r = 0.90) and a lower proportion of milk (UCOW, r = −0.85; UHA, r = −0.50) produced from grass or grass silage. Farms described by factor 1 also produced a large proportion of their milk between April and June (Yield46, r = 0.43), calved a significant proportion of cows in late summer/early autumn (CA79, r = 0.43) and had low margins per litre (MOFPLSTA, r = −0.45). 3.1.2. Factor 2 (milk yield and composition) Twelve variables loaded into factor 2 including milk yield (YIELD, r = 0.90, YIELDHA, r = 0.40), milk protein and fat production (YP, r = 0.92; YF, r = 0.90) and milk fat and protein percentage (PROT, r = 0.41; FAT, r = 0.30). Margins over feed and fertiliser were high (MOFPCSTA, r = 0.75; MOFHASTA, r = 0.42) and a significant proportion of milk per cow was produced from grass or grass silage (UCOW, r = 0.43) in early summer (Yield46, r = 0.46). Mean calving date was early (DAYC, r = −0.32; CA46, r = −0.36).
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
143
Fig. 1. Data structure revealed by varimax rotation. Variables with most significant loading for each factor are represented by bullet point.
144
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
3.1.3. Factor 3 (herd size) Factor 3 described large farms (HA, r = 0.93), with a large number of cows calving (NOCA, r = 0.97), a high total milk production (TOTMK, r = 0.97) and a large number of cows in the herd (HS, r = 0.96). Cows represented a high proportion of the total livestock units on the farm (CTOTLU, r = 0.34) and these farms also used a low proportion of commercial AI (COM, r = −0.37). 3.1.4. Factor 4 (stocking rate) Farms described by factor 4 had high stocking rates (SRHA, r = 0.92) produced a large amount of milk per hectare (YIELDHA, r = 0.85) and had high margins over feed and fertiliser per hectare (MOFHASTA, r = 0.81). A significant proportion of milk per hectare was produced from grass or grass silage (UHA, r = 0.81). 3.1.5. Factor 5 (summer-calving/milk price) Factor 5 described herds receiving a high milk price (MPSTAND, r = 0.84) and having a high margin over feed and fertiliser (MOFPLSTA, r = 0.75; MOFPCSTA, r = 0.54). These herds calved in late summer/early autumn (CA79, r = 0.60; CA16, r = −0.35; CA13, r = −0.42), had high calf mortality (CARIP, r = 0.32) and long lactation lengths (INMILK, r = 0.44). 3.1.6. Factor 6 (concentrate feeding and milk composition) Farms described by factor 6 fed low cost concentrates (CCSTAND, r = −0.75), fed concentrates to dry cows (MPCD, r = 0.53) and produced milk with high fat and protein concentrations (FAT, r = 0.69; PROT, r = 0.44). Mean calving date was late (DAYC, r = 0.43), a large proportion of cows calved between April and June (CA46, r = 0.54) while a relatively small amount of milk was produced between April and June (Yield46, r = −0.53). 3.1.7. Factor 7 (nitrogen and sulphur fertiliser) The variable with the highest loading in factor 7 was nitrogen fertiliser. Total units used per hectare (NHA, r = 0.84) and usage when grass silage and stocking rate were taken into account (RNHA, r = 0.84) were the most significant variables. This indicated that farms described by this factor used excess nitrogen. A high proportion of the nitrogen fertiliser usage occurred between March and May (NHA35, r = 0.69). These farms also applied a large amount of sulphur fertiliser to grassland (SULHA, r = 0.62). 3.1.8. Factor 8 (phosphorus and potassium fertiliser) Factor 8 relates exclusively to phosphorus (PHA, r = 0.96) and potassium (KHA, r = 0.95) fertiliser usage. 3.1.9. Factor 9 (milk quality) Farms with high somatic cell counts (SCC, r = 0.83) and high total bacteria counts (TBC, r = 0.76) were characterised by factor 9. These herds tended to calve cows in winter (CA16, r = −0.32; CA46, r = −0.35; Bosday, r = −0.32; CA1012, r = 0.43).
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
145
3.1.10. Factor 10 (spring-calving/high replacement rates) Factor 10 identified spring-calving herds (Bosday, r = 0.30; CA16, r = 0.31; CA1012, r = −0.36) that used commercial AI (COM, r = 0.34) and culled or sold a high proportion of cows (CULLPC, r = 0.34). As a result these herds had a high number of cows calving as percentage of the total cows in the herd (COCAPC, r = 0.83). 3.1.11. Factor 11 (young herds/high replacement rates) Herds described by factor 11 had young cows (LN, r = −0.51), high replacement rates (CULLPC, r = 0.44) and prolonged calving to first service intervals (CSE, r = 0.73). These herds also had high milk protein percentage (PROT, r = 0.35), high calf mortality (CARIP, r = 0.36) and a long lactation length (INMILK, r = 0.34). 3.1.12. Factor 12 (silage cut) The only significant variable that loaded into factor 12 was grass silage cut (CUT, r = 0.84). This indicated that farms classified by this factor cut a large proportion of their farm for grass silage. 3.1.13. Factor 13 (early spring-calving/dairy enterprise) Factor 13 related to herds calving cows in early spring (CA46, r = −0.33; CA79, r = −0.30) with low calf mortality (CARIP, r = −0.30), low replacement rates (CULLPC, r = −0.30) and that used commercial AI (COM, r = 0.37). These farms were predominantly dairy in their structure as cows formed a high percentage of total livestock units (COTOTLU, r = 0.64). 3.2. Regression (factors) Following backward elimination of non-significant factors, only one factor (factor 3 labelled herd size) was significantly (P < 0.01) associated with calving rate. The R2 value for this model was 40% and the coefficient of variation was 13.1%, while the adjusted R2 value was 27%. As indicated by the parameter estimates (±S.E.) B = −6.6 ± 2.6, for every one unit increase in the factor herd size, there was a 6.6% decrease in calving rate. The factor scores for factor 3 (herd size) were plotted against calving rate and because the plot was not linear it was decided that dividing the factor scores into quartiles would produce a better fitting model. The factor scores for factor 3 were sorted and assigned to four equal categories (n = 47 per category), from lowest to highest. The range in herd size according to category was as follows: category 1 (34–96), category 2 (47–103), category 3 (66–152) and category 4 (108–359). The calving rate (%), (±S.E.) was 67.0 ± 2.5 for category 1, 61.8 ± 1.8 for category 2, 56.9 ± 1.5 for category 3 and 53.2 ± 2.85 for category 4. Using pair-wise comparisons, calving rates differed significantly (P < 0.05) between all categories except between categories 3 and 4 (P > 0.05). Using the model, which included the categorised factor 3 instead of the factor scores for factor 3, there was a significant association (P < 0.01) with calving rate. The R2 value for this model was 42% while the adjusted R2 value was 26%. The original model used in this study was somewhat restrictive in that a large amount of the variation in calving rates was accounted for by inclusion of farm code as a random effect
146
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
in the model. A model fitted with farm code only resulted in an R2 of 34% with an adjusted R2 of 22%. If farm code was not included in the original model, four factors (factor 2, milk yield composition; factor 3, herd size; factor 6, concentrate feeding/milk composition and factor 10, spring-calving/high replacement rates) were negatively associated with calving rate. However, predictive ability of this model was low (R 2 = 22%; adjusted R 2 = 17%) in comparison to the predictive ability of the model containing year, farm code and factor scores for factor 3 (R 2 = 40%; adjusted R 2 = 27%). 4. Discussion Considerable research has examined and described dairy herd reproductive performance both in Ireland and in other countries. The traditional approach has been to use multiple regression in analysing the relationship between variables, such as milk yield and calving rate. This approach has limitations due to multicollinearity (i.e. where there are high correlations between predictor variables). An alternative approach uses factor analysis which was devised by two psychologists, Francis Galton and Charles Spearman in order to gain a better understanding of ‘intelligence’ (Chatfield and Collins, 1989). It was developed primarily for analysing relationships among a number of measurable entities, such as survey items or test scores. It is only relatively recently that factor analysis has been applied in animal and veterinary science studies. The main application of factor analysis is to reduce the number of variables to the smallest number that captures as much information as possible from the original dataset and to detect structure in the relationship between variables. The underlying assumption of factor analysis is that there exists a number of unobserved latent factors which account for the correlations among observed variables. Calving rate was selected as the outcome of interest in this study as it can be readily calculated using a consistent method in herds with different calving patterns and does not require pregnancy diagnosis results. However, using calving rate does have some limitations as it is independent of submission patterns and as such, does not fully describe herd reproductive performance. Furthermore, calving rate is susceptible to selection bias due to selective culling before subsequent calving of non-conceiving or late conceiving cows. Accordingly, calving rates with culled cows excluded from calculations (as done in this study) can be higher than conception rates when assessed by pregnancy diagnosis before substantial culling has occurred. If larger herds are more selective in their culling (i.e. a higher proportion of culls are not pregnant) than smaller herds, observed calving rates will be lower in larger herds, all else being equal. As pregnancy diagnosis information was not available in this study, it was not possible to remove this potential source of bias. There is a strong curvilinear relationship between calving to service interval and conception rate. Accordingly herds with many short intervals from calving to service will have lower conception (and calving rates), all else being equal. Therefore, in an attempt to control for the possible confounding of this effect the calving to service interval and the proportion of services within 60 days of calving were included in the final regression model to test these effects. However, when included, mean herd calving to service interval or proportion of services less than 60 days post-calving were not significant (P > 0.05) predictors of calving rate.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
147
In the current study, the factor analysis model identified a total of 13 factors with eigenvalues greater than one, indicating that the factors explained more of the variance in the original data than would a single variable by itself. Following backward elimination of non-significant factors, only factor 3 (herd size) was significantly (P < 0.01) associated with calving rate. Although factor 3 is labelled herd size, it includes variables other than the number of cows in the herd such as total milk production, farm size, number of cows calving, cows as a percent of total livestock units and usage of commercial AI. Although many of these variables appear to be similar, it is the linear combination of these variables that results in the factor labelled herd size and the associated negative effect on calving rate. However, because of the similarities between these variables and the strong correlation (r = 0.96, P < 0.001) between the variable herd size and the factor herd size, we concluded that the categories 1, 2, 3 and 4 equate to small, small to medium, medium to large and large herds. This indicates that larger herds have lower calving rates than medium or small herds. Eneveloldsen et al. (1996) included all two-factor interactions and the number of cows squared in a model to examine the importance of management types in herd milk production. For comparative purposes, all factor by year interactions and the variable, cows in the herd, were included as independent variables in an initial full regression model in the present study. There were significant interactions between years and factors 1, 2 and 9. Predictive ability of the model as indicated by the R2 values, increased from 40% (adjusted R 2 = 27%) to 58% (adjusted R 2 = 38%) when the interaction terms were included. However, it was difficult to interpret the biological significance of these interactions and it was decided to use a simple model, without the interaction terms. The biological interpretation of the effect of factor 3 on calving rate may relate to deficiencies in breeding management. Larger herds may have poorer oestrus detection due to reduced observation time, inadequate data management or reduced labour units per cow. Oestrus detection efficiency generally has declined in recent years as herd size and milk production have increased (Stevenson, 2001). The management approaches required to efficiently run large herds and the negative effects of herd size on herd health also potentially contribute to poor reproductive performance (Lucy, 2001). As reviewed by Nebel and McGilliard (1993), an important similarity in milk yield and reproductive efficiency is that management and environment account for the majority of the variation. The lack of an association between herd-level milk production and calving rate in the present study is an interesting finding. Low phenotypic correlations between milk production and fertility traits have been reported previously (Morton, 2000; Buckley et al., 2000). However, Buckley et al. (2000) did report an antagonistic genetic correlation between milk production and reproductive performance in seasonal calving systems. As a reliable genetic assessment for milk production was unavailable for most of the cows in the current study, it was not possible to remove this source of variation from the observed variability in calving rate between herds. However, milk yield per cow is most likely a proxy variable for genetic differences between herds. These results indicate that management of high producing herds may be sufficient to overcome adverse effects of high milk production on reproductive performance. Disease status, calving difficulty or lameness was not monitored in detail and could not be removed as a source of variation in the current study. Incidence of leptospirosis,
148
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
salmonellosis, brucellosis, infectious bovine rhinotracheitis and bovine viral diarrhoea virus are all likely to negatively impact on calving rate. Recent work in Moorepark (J. Crilly, personal communication) indicates that herd size is a significant predictor of the incidence of salmonellosis in herds. Reproductive performance is often reduced following problems at or after calving, including dystocia, twins, retained foetal membranes and vaginal discharge (Morton, 2000). Lameness can be attributed to many factors including nutrition, genetics and cow facilities, and particularly cubicle numbers per cow. Concrete yards and cow pathways, where animals have to travel long distances to pasture, are often inadequate where originally designed for smaller herds. Lameness during early lactation can have substantial effects on reproductive performance, particularly among cows suffering severe lameness during the first 6 weeks of mating and these effects are also associated with decreased submission rates (Morton, 2000). The financial implications of lower calving rates and delays in conception have been detailed recently by Esslemont et al. (2001). They concluded that there is considerable extra profit available from relatively small improvements in fertility management, but that the concept of lost profit due to long calving intervals and high culling rates needs to be explained and promoted to farmers.
5. Conclusion It is well documented that reproductive performance has declined in dairy herds and that this decline has been in association with dramatic increases in milk production. Based on the results of this study, some of that temporal decline is likely to be due to increased herd size over the same period. Herd-level milk production was not associated with calving rate in the present study which may indicate that any adverse effects of high milk production on reproductive performance can be overcome by good management. Due mainly to economies of scale, large herd sizes are likely to be a significant feature of dairy farming systems in the future. Large herd sizes may contribute to reduced reproductive performance and may require new approaches to breeding management (Lucy, 2001). Data from the present study support this conclusion indicating that larger herds, in combination with other associated herd characteristics, are likely to have poorer calving rates than medium or small herds.
Acknowledgements The authors wish to acknowledge D. Cliffe for technical assistance and the farmers who participated in DairyMIS. References ACITS, 1995. Factor analysis using SAS Proc factor. Statistical services, University of Texas, Austin, USA. Available at: http://www.Utexas.edu/cc/docs/stat53.html. Accessed 6 April 2001. Buckley, F., Dillon, P., Mee, J.F., Veerkamp, R., 2000. Moorepark fertility project—latest results. In: Proceedings of Irish Grassland Association Dairy Conference. Tullamore, Ireland, pp. 24–37.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
149
Chatfield, C., Collins, A.J., 1989. Introduction to Multivariate Analysis. Chapman & Hall, London. Crosse, S., 1985. Dairy Herd Management Information System—DairyMIS II. Part 2: Animal Health and Machine Milking. Moorepark 25th Anniversary Publication, pp. 325–361. Crowley, J.P., Harrington, D., Lacy, M., 1967. A survey of reproductive efficiency in cattle. Ir. J. Agric. Res. 6, 237–246. Cunningham, E.P., O’Byrne, T.M., Murphy, N., 1978. Survey of A.I. Results 1978: Final Report Trinity College Dublin. Applied Research and Consultancy Group, Dublin 2, Ireland. Cunningham, E.P., O’Byrne, T.M., 1980. Survey of A.I. Results 1980: Final Report Trinity College Dublin. Applied Research and Consultancy Group, Dublin 2, Ireland. Dillon, P., 1996. Maximising profit in creamery milk production using current research and technology. Ir. Grassland Anim. Prod. Assoc. J. 30, 8–21. Eneveloldsen, C., Sorensen, J.T., Thysen, I., Guard, C., Grohn, Y.T., 1995. A diagnostic and prognostic tool for epidemiologic and economic analyses of dairy herd health management. J. Dairy Sci. 78, 947–961. Eneveloldsen, C., Hindhede, J., Kristensen, T., 1996. Dairy herd management types assessed from indicators of health, reproduction, replacement and milk production. J. Dairy Sci. 79, 1221–1236. Esslemont, R.J., Kossaibiti, M.A., Allcock, J., 2001. Economics of fertility in dairy cows. In: Fertility in the High-Producing Dairy Cow, vol. 26. British Society of Animal Science Occasional Publication, pp. 19–29. Funk, D.A., 1993. Optimal genetic improvement for the high producing herd. J. Dairy Sci. 76, 3278–3286. Gilbert, R.P., Bailey, D.R.C., 1991. Hair coat characteristics and postweaning growth of Hereford and Angus cattle. J. Anim. Sci. 69, 498–506. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data Analysis. Prentice Hall, Upper Saddle River, New Jersey. Hoekstra, J., Van der Lugt, A.W., Van der Werf, J.H.J., Ouweltjes, W., 1994. Genetic and phenotypic parameters for milk production and fertility traits in upgraded dairy cattle. Livest. Prod. Sci. 40, 225–232. Holloway, J.W., Savell, J.W., Hamby, P.L., Baker, J.F., Stouffer, J.R., 1990. Relationships of empty-body composition and fat distribution to live animal and carcass measurements in Bos Indicus-Bos Taurus crossbred cows. J. Anim. Sci. 68, 1818–1826. Hurnik, D., Dohoo, I.R., Donald, A., Robinson, N.P., 1994. Factor analysis of swine farm management practices on Prince Edward Island. Prev. Vet. Med. 20, 135–146. Johnson, R.A., Wichern, D.W., 1982. Applied Multivariate Statistical Analysis. Prentice Hall, Englewood Cliffs, New Jersey. Lucy, M.C., 2001. Reproductive loss in high-producing dairy cattle: where will it end? J. Dairy Sci. 84, 1277–1293. Mee, J.F., Fahey, J., Crilly, J., 1999. Breeding the dairy cow of the future—todays challenges. In: Proceedings of the Teagasc National Dairy Conference on Dairying in the New Millennium. Adare, Teagasc, Moorepark, Fermoy, Co. Cork, Ireland, pp. 7–16. Morton, J.M., 2000. The in-calf project—some risk factors for reproductive performance in Australian dairy herds. In: Proceedings of the Australian and New Zealand Combined Dairy Veterinarians’ Conference. Le Lagon Resort, Port Vila, Vanuatu, Australia, ISSN 0112-9643, pp. 43–61. Nebel, R.L., McGilliard, M.L., 1993. Interactions of high milk yield and reproductive performance in dairy cows. J. Dairy Sci. 76, 3257–3268. O’Farrell, K.J., Crilly, J., 1998. Fertility rates in Irish dairy herds. In: Proceedings of the 20th World Buiatrics Congress, vol. 2. Sydney, Australia, pp. 585–587. Price, E.O., Wallach, S.J.R., 1991. Sexual and aggressive behaviours of Hereford bulls: interrelationship of variables and problems with the use of composite scores to describe overall sexual performance. J. Anim. Sci. 69, 1028–1033. Pryce, J.E., Esslemont, R.J., Thompson, R., Veerkamp, R.F., Kossaibati, M.A., Simm, G., 1998. Estimation of genetic parameters using health, fertility and production data from a management recording system for dairy cattle. Anim. Sci. 66, 577–584. Roche, J.F., Sherington, J., Mitchell, J.P., Cunningham, J.F., 1978. Factors affecting calving rate to A.I. in cows. Ir. J. Agric. Res. 17, 149–157. SAS User’s Guide: Statistics, 1989. Version 6, 4th ed., vol. 2. SAS Institute, Cary, NC. Shahin, K.A., Ashmawy, A.A., Mourad, K.A., Foda, T.A., 1998. Varimax rotated factors derived from body dimensions as used to predict production criteria in buffalo cows. Indian J. Dairy Sci. 68, 946–949.
150
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Sieber, M., Freeman, A.E., Hinz, P.N., 1987. Factor analysis for evaluating relationships between first lactation type scores and production data of Holstein dairy cows. J. Dairy Sci. 70, 1018–1026. Stevenson, J.S., 2001. A review of oestrus behaviour and detection in dairy cows. In: Fertility in the High-Producing Dairy Cow, vol. 26. British Society of Animal Science Occasional Publication, pp. 43–62. Van Arendonk, J.A.M., Nieuwhof, G.J., Vos, H., Korver, S., 1991. Genetic aspects of feed intake and efficiency in lactating dairy heifers. Livest. Prod. Sci. 29, 263–275. White, D.S., O’Farrell, J.D., 1972. The artificial insemination service and conception rates. Vet. Services Bull., UK 1, 60–66.
The effect of feeding and management practices on calving rate in dairy herds J. Fahey a,1 , K. O’Sullivan b , J. Crilly a , J.F. Mee a,∗ a
b
Teagasc, Dairy Production Department, Dairy Production Research Centre, Moorepark, Fermoy, Co. Cork, Ireland Statistical Laboratory, Department of Statistics, University College Cork, Cork, Ireland
Received 29 November 2001; received in revised form 15 May 2002; accepted 23 August 2002
Abstract The objective of this study was to examine the effects of nutrition and management practices on reproductive performance in 31 Irish dairy herds participating in the Moorepark Dairy Management Information System (DairyMIS) during the period 1991–1998. Fifty variables relating to herd reproductive indices, calving events, stocking rate, disease, concentrate feeding, fertiliser usage, milk production and economic performance were studied using factor analysis. A factor analysis, followed by varimax rotation, identified 13 factors, which accounted for 83% of the total variance of the original variables. A regression model was used to predict calving rate from the orthogonal factor scores identified by factor analysis. Calving rate was defined as the proportion of services, for which an outcome was known, which resulted in a subsequent calving. Year, farm code and factor 3 (labelled herd size) together accounted for 40% of the observed variation in calving rate. The factor scores for factor 3 (herd size) were plotted against calving rate and because the plot was not linear, it was decided that dividing the factor scores into quartiles would produce a better fitting model. The factor scores for herd size were sorted and assigned to four equal categories (n = 47 per category), from lowest to highest. The ranges in herd size according to category were as follows: category 1 (34–96), category 2 (47–103), category 3 (66–152) and category 4 (108–359). The calving rate (%), (±S.E.) was 67.0 ± 2.5 for category 1, 61.8 ± 1.8 for category 2, 56.9 ± 1.5 for category 3 and 53.2 ± 2.85 for category 4. Using pair-wise comparisons, calving rates differed significantly (P < 0.05) between all categories except between categories 3 and 4 (P > 0.05). Herd-level milk production was not associated with calving rate indicating that good management
∗ Corresponding author. Tel.: +353-25-42387; fax: +353-25-42340. E-mail address: [email protected] (J.F. Mee). 1 Present address: Department of Veterinary Science, Victorian Institute of Animal Science, University of Melbourne, 600 Sneydes Road, Werribee, Victoria 3030, Australia.
0378-4320/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 4 3 2 0 ( 0 2 ) 0 0 1 9 2 - 6
134
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
may overcome any adverse effects of high milk production on reproductive performance. Larger herds, in combination with other associated herd characteristics, are likely to have poorer calving rates than medium or small herds. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Factor analysis; Fertility; Herd size; Nutrition; Management; Dairy cows
1. Introduction Optimising reproductive efficiency is essential in maintaining profitability in a dairy herd. The economic advantages of a compact calving pattern in seasonal milk production, which allows 80% of milk to be produced from spring grass, have been quantified by Dillon (1996). While genetic improvement of dairy cows has led to a dramatic increase in milk yield (Funk, 1993), genetic correlation estimates indicate an antagonistic relationship between milk production and fertility (Van Arendonk et al., 1991; Hoekstra et al., 1994; Pryce et al., 1998). Reproductive performance of Irish dairy herds has been assessed previously using research databases (O’Farrell and Crilly, 1998; Mee et al., 1999; Buckley et al., 2000) and surveys (Crowley et al., 1967; White and O’Farrell, 1972; Roche et al., 1978, Cunningham et al., 1978; Cunningham and O’Byrne, 1980). In the studies undertaken between 1967 and 1980, calving rate to first service was reported to be 60–69%. In the study reported by Mee et al. (1999), calving rate to first service fell significantly (0.9% per year) from 59% in 1991 to 54% in 1998. The potential effects of nutritional and management factors and their changes over time on reproductive performance in spring-calving Irish dairy herds have not been examined previously. Evaluation of management across a large number of herds can be difficult. Data may be incomplete and not comparable. A dairy herd is a complex and dynamic system in which input and output are related to management strategies and animal status in a complex manner (Eneveloldsen et al., 1995). Many management indicators are strongly interrelated making it difficult to specify explicitly which variables are dependent and which are independent in a traditional analysis (Eneveloldsen et al., 1996). Data reduction techniques, such as principal component analysis and factor analysis, have been successfully used to evaluate underlying relationships among multiple traits or management types measured simultaneously (Sieber et al., 1987; Holloway et al., 1990; Price and Wallach, 1991; Hurnik et al., 1994; Eneveloldsen et al., 1996; Shahin et al., 1998). The objective of this study was to establish the association between dairy herd management indicators (calving events, stocking rate, milk yield and milk composition, concentrate feeding, fertiliser usage, disease and economic performance) and calving rate in herds representative of seasonal milk production systems and participating in the Moorepark Dairy Management Information System (DairyMIS) during the period 1991– 1998.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
135
2. Materials and methods 2.1. Data Information on farm management practices was compiled from the DairyMIS database as described by Crosse (1985). The database consisted of 32 farms (herds) recorded over an 8-year period from 1991 to 1998 (284 herd years) although some of the farms were not present for the entire 8-year period. Calving rate (CR) was defined as the proportion of services, for which an outcome was known, which resulted in a subsequent calving. Where a service was followed by a subsequent calving at an interval of no more than 300 days, the animal was deemed to have conceived to that service. Where a service was followed by a further service event at an interval of at least 5 days, the initial service was considered to have failed. Where two services were recorded within an interval of 4 days the second service was ignored and any conception was attributed to the first event. If a calving event occurred more than 300 days after the last recorded service event, the cow was assumed to have conceived to an unrecorded service (O’Farrell and Crilly, 1998). If greater than 10% of cows in a given year calved to an unrecorded service the records were considered to be unreliable and the data for that herd-year were excluded from the analysis. Where more than 30% of cows served in a given year failed to calve in the following year the data for that herd-year were again excluded from the analysis. Very low calving rates tended to be associated with events such as partial depopulations following outbreaks of notifiable diseases and their inclusion in the database could be a major source of bias. When herd-years were excluded as described, the dataset consisted of 31 herds and 188 herd-years. The participating farms were commercial dairy herds with an average of 108 cows (range 34–359) producing an average milk yield of 5000 kg per cow per year (range 4000–7000 kg). The herds were predominantly spring/summer-calving with more than three-quarters of the herds having ≥85% of their cows calving between January and June. The farms involved in the study were situated in counties Cork, Waterford and Tipperary in southern Ireland. 2.2. DairyMIS variables The variables (all continuous) recorded in the DairyMIS database relate to herd reproductive indices, calvings, stocking rate, disease, concentrate feeding, fertiliser usage, milk production and economic indices (Tables 1–8). Herd-year was used as the unit of analysis. The name, abbreviation (e.g. stocking rate per hectare (name), SRHA (abbreviation)) and definition for the 50 variables in the database are presented. 2.3. Statistical analyses 2.3.1. Factor analysis A common factor analysis model using the Proc Factor procedure of SAS (1989) computer software was used to condense the information contained in the 50 original variables into a smaller set of dimensions (factors). The correlation matrix of each variable with all other variables was the basis for factor extraction. The variance accounted for by each factor,
136
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 1 Variables relating to herd reproductive indices recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Calving to first service interval (CSE)
The mean interval in days from last calving date to first service date. The proportion of artificial inseminations carried out by a commercial technician. It was possible to categorise inseminations as commercial, DIY or natural service based on information supplied by the farmer, together with inseminator codes. The date of the first service event in days of year (1–365). If no further cows were served within 5 days of the first service event then the first service event was ignored and the next service date was taken as the beginning of the breeding season provided two further service events occurred within a 10-day period. In order to overcome the calendar year-on-year effect (i.e. 31 December = day 365 and 1 January = day 1), when beginning of the breeding season day was >200 it was taken as that day of year minus 365. This gave a continuous breeding season from July to July.
Commercial AI (COM)
Beginning of breeding season (Bosday)
its eigenvalue, was used to determine its significance and whether it would be retained for further analysis (Hair et al., 1998). Because the original factor loadings (the correlations between the underlying factors and the original variables) were not readily interpretable, an orthogonal transformation matrix Table 2 Variables relating to cows calving recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total number of calvings (NOCA) Herd size (HS) Cows calving percentage (COCAPC) Mean calving date (DAYC)
Total number of cows calving Total number of cows in the herd Number of calvings/cows in herd × 100 If calving day was greater than day 200 then mean calving day was set to calving day minus 365. Similar to beginning of breeding season day this was done to take into account the difference between 31 December (day 365) and 1 January (day 1) where cows calve from one year to the next. Number of cows calving from January to June/total number of calvings Number of cows calving from January to March/total number of calvings Number of cows calving from April to June/total number of calvings Number of cows calving from July to September/total number of calvings Number of cows calving from October to December/total number of calvings Cows in milk/cows in herd × 100
Cows calving from January to June (CA16) Cows calving from January to March (CA13) Cows calving from April to June (CA46) Cows calving from July to September (CA79) Cows calving from October to December (CA1012) Cows in milk percentage, lactation length (INMILK)
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
137
Table 3 Variables relating to stocking rate recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Culling/replacement rate (CULLPC) Number of hectares (HA) Stocking rate/HA (SRHA)
Number of cows sold/cows in herd × 100 Number of hectares allocated to the dairy enterprise Area of land devoted to dairy enterprise (HA)/total livestock units (number of cows multiplied by 1, plus heifers in calf multiplied by 0.7, plus cattle under 1 year multiplied by 0.3, plus cattle between 1 and 2 years multiplied by 0.7, plus cattle over 2 years multiplied by 1, plus cull cows multiplied by 1, plus any other livestock units (horses, sheep, deer) Cows in herd/total livestock units × 100
Cows as percentage of total livestock units (COTOTLU) Lactation number (LN) Silage cut (Cut)
Average number of lactations for cows in herd Percentage of farm cut for first + second + third cut grass silage
Table 4 Variables relating to disease recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Calf mortality (CARIP)
Number of births dead/births total × 100. Number of births dead includes certain abortion, uncertain abortion, stillbirth, dead in 24 h, dead in 48 h and dead between 3 days and 3 weeks Taken from milk test (weighted average over time) Taken from milk test (weighted average over time)
Somatic cell count (SCC) Total bacteria count (TBC)
Table 5 Variables relating to concentrate feeding recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Concentrate fed per cow per year (CPC) Residual concentrate feeding per year (RMEAL)
Total tonnes of concentrates used/cows in herd Residual concentrate feeding when milk yield was taken into account The residuals are calculated from the model: CPC = constant + yield. Concentrate fed per cow/(cows in herd−cows in milk) Concentrate fed per cow/cows in milk Concentrate fed per cow/litres per cow Utilised metabolisable energy (UME)/HA. UME was calculated assuming that 25,000 MJ of energy was required per cow per year for maintenance, 5 MJ of energy was required to produce a litre of milk, and concentrates provide 11.5 MJ/kg UME/cows in herd
Concentrate fed/cows dry per year (MPCD) Concentrate fed/cows in milk per year (MPCIM) Concentrate per litre per year (MPL) Utilised metabolisable energy/HA per year (UHA)
Utilisable metabolisable energy per cow per year (UCOW)
138
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 6 Fertiliser variables recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total units of nitrogen/HA per year (NHA) Total units of nitrogen/HA from March to May (NHA35)
Units of nitrogen fertiliser used/HA Units of nitrogen fertiliser used from March to May/HA Residual units of nitrogen fertiliser/HA. The residuals are calculated from the model: NHA = constant + SRHA + Cut. Units of phosphorus fertiliser/HA Units of potassium fertiliser/HA Units of sulphur fertiliser/HA
Residual units of nitrogen/HA per year (RNHA)
Total units of phosphorus fertiliser/HA per year (PHA) Total units of potassium fertiliser/HA per year (KHA) Total units of sulphur fertiliser/HA per year (SULHA)
Table 7 Milk production variables recorded in DairyMIS including name, abbreviation and definition Variable
Definition
Total milk produced per year (TOTMK) Milk fat yield in litres per year (YF) Milk protein yield in litres per year (YP) Milk fat percentage per year (FAT) Milk protein percentage per year (PROT) Milk production per cow in litres per year (Yield) Yield of milk/HA per year (Yieldha) Yield of milk between April and June (Yield46)
Total herd milk production Total milk fat production in litres Total milk protein production in litres Milk fat percentage Milk protein percentage Total milk produced/cows in herd Milk production per cow in litres × stocking rate/HA Total milk produced between April and June/cows in milk between April and June
Table 8 Economic indices recorded in DairyMIS including variable name, abbreviation and definition Variable
Definition
Concentrate cost per year (CCSTAND) Milk price (MPSTAND)
Concentrate cost is equal to the total tonnes used × cost per tonne. Milk price standardised within year (Concentrate cost per tonne × total concentrates fed)/cows in herd + (other feed costs per cow) (Concentrate cost per tonne × total concentrates fed)/cows in herd + (other feed costs per cow)/milk yield per cow (Milk value−concentrate cost−other feed costs−fertiliser cost)/cows in herd (Milk value−concentrate cost−other feed costs−fertiliser cost)/HA Milk value−concentrate cost−other feed costs−fertiliser cost/yield
Feed costs per cow per year (FCCSTAND)
Feed costs per litre per year (FCLSTAND)
Margin over feed and fertiliser per cow per year (MOFPCSTA) Margin over feed and fertiliser/HA per year (MOFHASTA) Margin over feed and fertiliser per litre per year (MOFPLSTA)
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
139
140
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
141
was used to rotate the factor loadings (Hair et al., 1998). Rotation using an orthogonal transformation matrix (e.g. varimax rotation) maintains the independence of the factors (Gilbert and Bailey, 1991). After extraction of the initial factor solution (common factors), the matrix was rotated, using the varimax method (Kaiser normalisation), in order to obtain a clear pattern of loadings, thereby making the results more interpretable (Table 9). Each variable then had a high loading on a single factor and a small to moderate loading on other factors (Johnson and Wichern, 1982). When assessing factors, it is important that they make biological sense. Factors may be rotated by several techniques to simplify the factor structure and facilitate interpretation. If more than one common factor is identified in a factor analysis and an oblique rotation has produced correlations between the factors, the interrelations between these first-order factors can be described by a second factor analysis based on factor scores or interfactor correlations from the first-order analysis (Eneveloldsen et al., 1996). The second-order factor analysis then identifies a second set of common factors that explains the variation in the obliquely rotated first-order factors. As the study reported by Eneveloldsen et al. (1996) was similar in its data and objectives, second-order factor analysis was attempted in the current study. Four second-order factors were produced. However, these factors were uninterpretable, therefore, first-order factor analysis using orthogonal rotation was preferred. In the present study, the factors were labelled according to the pattern of factor loadings. Significant loadings for each factor were examined and, in general, the larger the absolute size of the factor loading for a variable, the more important the variable was in interpreting the factor. All the variables loading on a factor were considered including the size and sign of the loading. A determination was then made as to what the underlying factor represented (ACITS, 1995). 2.4. Regression analysis (factors) The scores of the 13 factors were used as determinants in a general linear model in order to identify which, if any, of the factors relating to feeding and farm management practices were associated with calving rate. Calving rate to all services was taken as the variable, which best represented herd reproductive performance. Calving rate was used as the dependent variable in a linear regression model (original model) which contained year, farm code and the scores for each of the 13 factors as determinants. Farm code was classified as a random effect and year as a fixed effect. The regression procedure utilised was backward elimination. Initially all factor scores were included in the model and the least significant factor scores were removed one at a time until each factor score in the model had an overall probability of <0.05. Once eliminated, factor scores were not allowed to be included again in the model building process. Goodness of fit was evaluated from residual plots and predictive ability was assessed by R2 values. Residual diagnostics did not indicate any concern for departures from the statistical assumptions of constant variability and normality. 3. Results The mean calving rate to all services across all herd-years was 58% while the lowest and highest herd-year calving rates ranged from 35 to 85%. The range in calving rate between
142
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Table 10 Descriptive statistics of calving rate to all services, 1991–1998 Calving rate to all services
1991
1992
1993
1994
1995
1996
1997
1998
Number of herds Mean Standard error Median Minimum Maximum
18 0.59 0.03 0.60 0.35 0.85
25 0.60 0.02 0.59 0.42 0.81
26 0.58 0.02 0.57 0.36 0.74
27 0.56 0.01 0.55 0.43 0.72
24 0.60 0.02 0.62 0.42 0.76
20 0.59 0.02 0.58 0.46 0.70
24 0.55 0.02 0.55 0.41 0.72
24 0.57 0.02 0.58 0.41 0.73
herd averages across years was 49–74%. The mean range in calving rate to all services within herd between years was 18% where the lowest was 0% and the highest 38%. Descriptive statistics of calving rate across herds between years are presented (Table 10). 3.1. Factor analysis Thirteen factors, or latent variables of the initial factor solution, accounted for 83% of the total variance in the 50 variables. The rotated factor loading ‘r’ is similar in its interpretation to the correlation coefficient. Factor loadings equivalent to correlation coefficients of 0.30 and greater are highlighted to indicate the main attributes of the different factors (Table 9). The data structure between the 50 original variables and the 13 factors is presented in Fig. 1. 3.1.1. Factor 1 (winter-calving/concentrate feeding) Factor 1 was the most clearly defined, with 16 of the 50 original variables loading significantly. Farms with a high score for factor 1 had a high proportion of cows calving in the winter (CA1012, r = 0.60) and a low proportion calving in early spring (CA13, r = −0.63; CA16, r = −0.64). As a consequence, mean calving date (DAYC, r = −0.58) and start of the breeding season (BOSDAY, r = −0.65) were early. Concentrate feeding (MPL, r = 0.97; C50PC, r = 0.95; RMEAL, r = 0.95; MPCIM, r = 0.92) was high which resulted in high feed costs (FCLSTAND, r = 0.90; FCCSTAND, r = 0.90) and a lower proportion of milk (UCOW, r = −0.85; UHA, r = −0.50) produced from grass or grass silage. Farms described by factor 1 also produced a large proportion of their milk between April and June (Yield46, r = 0.43), calved a significant proportion of cows in late summer/early autumn (CA79, r = 0.43) and had low margins per litre (MOFPLSTA, r = −0.45). 3.1.2. Factor 2 (milk yield and composition) Twelve variables loaded into factor 2 including milk yield (YIELD, r = 0.90, YIELDHA, r = 0.40), milk protein and fat production (YP, r = 0.92; YF, r = 0.90) and milk fat and protein percentage (PROT, r = 0.41; FAT, r = 0.30). Margins over feed and fertiliser were high (MOFPCSTA, r = 0.75; MOFHASTA, r = 0.42) and a significant proportion of milk per cow was produced from grass or grass silage (UCOW, r = 0.43) in early summer (Yield46, r = 0.46). Mean calving date was early (DAYC, r = −0.32; CA46, r = −0.36).
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
143
Fig. 1. Data structure revealed by varimax rotation. Variables with most significant loading for each factor are represented by bullet point.
144
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
3.1.3. Factor 3 (herd size) Factor 3 described large farms (HA, r = 0.93), with a large number of cows calving (NOCA, r = 0.97), a high total milk production (TOTMK, r = 0.97) and a large number of cows in the herd (HS, r = 0.96). Cows represented a high proportion of the total livestock units on the farm (CTOTLU, r = 0.34) and these farms also used a low proportion of commercial AI (COM, r = −0.37). 3.1.4. Factor 4 (stocking rate) Farms described by factor 4 had high stocking rates (SRHA, r = 0.92) produced a large amount of milk per hectare (YIELDHA, r = 0.85) and had high margins over feed and fertiliser per hectare (MOFHASTA, r = 0.81). A significant proportion of milk per hectare was produced from grass or grass silage (UHA, r = 0.81). 3.1.5. Factor 5 (summer-calving/milk price) Factor 5 described herds receiving a high milk price (MPSTAND, r = 0.84) and having a high margin over feed and fertiliser (MOFPLSTA, r = 0.75; MOFPCSTA, r = 0.54). These herds calved in late summer/early autumn (CA79, r = 0.60; CA16, r = −0.35; CA13, r = −0.42), had high calf mortality (CARIP, r = 0.32) and long lactation lengths (INMILK, r = 0.44). 3.1.6. Factor 6 (concentrate feeding and milk composition) Farms described by factor 6 fed low cost concentrates (CCSTAND, r = −0.75), fed concentrates to dry cows (MPCD, r = 0.53) and produced milk with high fat and protein concentrations (FAT, r = 0.69; PROT, r = 0.44). Mean calving date was late (DAYC, r = 0.43), a large proportion of cows calved between April and June (CA46, r = 0.54) while a relatively small amount of milk was produced between April and June (Yield46, r = −0.53). 3.1.7. Factor 7 (nitrogen and sulphur fertiliser) The variable with the highest loading in factor 7 was nitrogen fertiliser. Total units used per hectare (NHA, r = 0.84) and usage when grass silage and stocking rate were taken into account (RNHA, r = 0.84) were the most significant variables. This indicated that farms described by this factor used excess nitrogen. A high proportion of the nitrogen fertiliser usage occurred between March and May (NHA35, r = 0.69). These farms also applied a large amount of sulphur fertiliser to grassland (SULHA, r = 0.62). 3.1.8. Factor 8 (phosphorus and potassium fertiliser) Factor 8 relates exclusively to phosphorus (PHA, r = 0.96) and potassium (KHA, r = 0.95) fertiliser usage. 3.1.9. Factor 9 (milk quality) Farms with high somatic cell counts (SCC, r = 0.83) and high total bacteria counts (TBC, r = 0.76) were characterised by factor 9. These herds tended to calve cows in winter (CA16, r = −0.32; CA46, r = −0.35; Bosday, r = −0.32; CA1012, r = 0.43).
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
145
3.1.10. Factor 10 (spring-calving/high replacement rates) Factor 10 identified spring-calving herds (Bosday, r = 0.30; CA16, r = 0.31; CA1012, r = −0.36) that used commercial AI (COM, r = 0.34) and culled or sold a high proportion of cows (CULLPC, r = 0.34). As a result these herds had a high number of cows calving as percentage of the total cows in the herd (COCAPC, r = 0.83). 3.1.11. Factor 11 (young herds/high replacement rates) Herds described by factor 11 had young cows (LN, r = −0.51), high replacement rates (CULLPC, r = 0.44) and prolonged calving to first service intervals (CSE, r = 0.73). These herds also had high milk protein percentage (PROT, r = 0.35), high calf mortality (CARIP, r = 0.36) and a long lactation length (INMILK, r = 0.34). 3.1.12. Factor 12 (silage cut) The only significant variable that loaded into factor 12 was grass silage cut (CUT, r = 0.84). This indicated that farms classified by this factor cut a large proportion of their farm for grass silage. 3.1.13. Factor 13 (early spring-calving/dairy enterprise) Factor 13 related to herds calving cows in early spring (CA46, r = −0.33; CA79, r = −0.30) with low calf mortality (CARIP, r = −0.30), low replacement rates (CULLPC, r = −0.30) and that used commercial AI (COM, r = 0.37). These farms were predominantly dairy in their structure as cows formed a high percentage of total livestock units (COTOTLU, r = 0.64). 3.2. Regression (factors) Following backward elimination of non-significant factors, only one factor (factor 3 labelled herd size) was significantly (P < 0.01) associated with calving rate. The R2 value for this model was 40% and the coefficient of variation was 13.1%, while the adjusted R2 value was 27%. As indicated by the parameter estimates (±S.E.) B = −6.6 ± 2.6, for every one unit increase in the factor herd size, there was a 6.6% decrease in calving rate. The factor scores for factor 3 (herd size) were plotted against calving rate and because the plot was not linear it was decided that dividing the factor scores into quartiles would produce a better fitting model. The factor scores for factor 3 were sorted and assigned to four equal categories (n = 47 per category), from lowest to highest. The range in herd size according to category was as follows: category 1 (34–96), category 2 (47–103), category 3 (66–152) and category 4 (108–359). The calving rate (%), (±S.E.) was 67.0 ± 2.5 for category 1, 61.8 ± 1.8 for category 2, 56.9 ± 1.5 for category 3 and 53.2 ± 2.85 for category 4. Using pair-wise comparisons, calving rates differed significantly (P < 0.05) between all categories except between categories 3 and 4 (P > 0.05). Using the model, which included the categorised factor 3 instead of the factor scores for factor 3, there was a significant association (P < 0.01) with calving rate. The R2 value for this model was 42% while the adjusted R2 value was 26%. The original model used in this study was somewhat restrictive in that a large amount of the variation in calving rates was accounted for by inclusion of farm code as a random effect
146
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
in the model. A model fitted with farm code only resulted in an R2 of 34% with an adjusted R2 of 22%. If farm code was not included in the original model, four factors (factor 2, milk yield composition; factor 3, herd size; factor 6, concentrate feeding/milk composition and factor 10, spring-calving/high replacement rates) were negatively associated with calving rate. However, predictive ability of this model was low (R 2 = 22%; adjusted R 2 = 17%) in comparison to the predictive ability of the model containing year, farm code and factor scores for factor 3 (R 2 = 40%; adjusted R 2 = 27%). 4. Discussion Considerable research has examined and described dairy herd reproductive performance both in Ireland and in other countries. The traditional approach has been to use multiple regression in analysing the relationship between variables, such as milk yield and calving rate. This approach has limitations due to multicollinearity (i.e. where there are high correlations between predictor variables). An alternative approach uses factor analysis which was devised by two psychologists, Francis Galton and Charles Spearman in order to gain a better understanding of ‘intelligence’ (Chatfield and Collins, 1989). It was developed primarily for analysing relationships among a number of measurable entities, such as survey items or test scores. It is only relatively recently that factor analysis has been applied in animal and veterinary science studies. The main application of factor analysis is to reduce the number of variables to the smallest number that captures as much information as possible from the original dataset and to detect structure in the relationship between variables. The underlying assumption of factor analysis is that there exists a number of unobserved latent factors which account for the correlations among observed variables. Calving rate was selected as the outcome of interest in this study as it can be readily calculated using a consistent method in herds with different calving patterns and does not require pregnancy diagnosis results. However, using calving rate does have some limitations as it is independent of submission patterns and as such, does not fully describe herd reproductive performance. Furthermore, calving rate is susceptible to selection bias due to selective culling before subsequent calving of non-conceiving or late conceiving cows. Accordingly, calving rates with culled cows excluded from calculations (as done in this study) can be higher than conception rates when assessed by pregnancy diagnosis before substantial culling has occurred. If larger herds are more selective in their culling (i.e. a higher proportion of culls are not pregnant) than smaller herds, observed calving rates will be lower in larger herds, all else being equal. As pregnancy diagnosis information was not available in this study, it was not possible to remove this potential source of bias. There is a strong curvilinear relationship between calving to service interval and conception rate. Accordingly herds with many short intervals from calving to service will have lower conception (and calving rates), all else being equal. Therefore, in an attempt to control for the possible confounding of this effect the calving to service interval and the proportion of services within 60 days of calving were included in the final regression model to test these effects. However, when included, mean herd calving to service interval or proportion of services less than 60 days post-calving were not significant (P > 0.05) predictors of calving rate.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
147
In the current study, the factor analysis model identified a total of 13 factors with eigenvalues greater than one, indicating that the factors explained more of the variance in the original data than would a single variable by itself. Following backward elimination of non-significant factors, only factor 3 (herd size) was significantly (P < 0.01) associated with calving rate. Although factor 3 is labelled herd size, it includes variables other than the number of cows in the herd such as total milk production, farm size, number of cows calving, cows as a percent of total livestock units and usage of commercial AI. Although many of these variables appear to be similar, it is the linear combination of these variables that results in the factor labelled herd size and the associated negative effect on calving rate. However, because of the similarities between these variables and the strong correlation (r = 0.96, P < 0.001) between the variable herd size and the factor herd size, we concluded that the categories 1, 2, 3 and 4 equate to small, small to medium, medium to large and large herds. This indicates that larger herds have lower calving rates than medium or small herds. Eneveloldsen et al. (1996) included all two-factor interactions and the number of cows squared in a model to examine the importance of management types in herd milk production. For comparative purposes, all factor by year interactions and the variable, cows in the herd, were included as independent variables in an initial full regression model in the present study. There were significant interactions between years and factors 1, 2 and 9. Predictive ability of the model as indicated by the R2 values, increased from 40% (adjusted R 2 = 27%) to 58% (adjusted R 2 = 38%) when the interaction terms were included. However, it was difficult to interpret the biological significance of these interactions and it was decided to use a simple model, without the interaction terms. The biological interpretation of the effect of factor 3 on calving rate may relate to deficiencies in breeding management. Larger herds may have poorer oestrus detection due to reduced observation time, inadequate data management or reduced labour units per cow. Oestrus detection efficiency generally has declined in recent years as herd size and milk production have increased (Stevenson, 2001). The management approaches required to efficiently run large herds and the negative effects of herd size on herd health also potentially contribute to poor reproductive performance (Lucy, 2001). As reviewed by Nebel and McGilliard (1993), an important similarity in milk yield and reproductive efficiency is that management and environment account for the majority of the variation. The lack of an association between herd-level milk production and calving rate in the present study is an interesting finding. Low phenotypic correlations between milk production and fertility traits have been reported previously (Morton, 2000; Buckley et al., 2000). However, Buckley et al. (2000) did report an antagonistic genetic correlation between milk production and reproductive performance in seasonal calving systems. As a reliable genetic assessment for milk production was unavailable for most of the cows in the current study, it was not possible to remove this source of variation from the observed variability in calving rate between herds. However, milk yield per cow is most likely a proxy variable for genetic differences between herds. These results indicate that management of high producing herds may be sufficient to overcome adverse effects of high milk production on reproductive performance. Disease status, calving difficulty or lameness was not monitored in detail and could not be removed as a source of variation in the current study. Incidence of leptospirosis,
148
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
salmonellosis, brucellosis, infectious bovine rhinotracheitis and bovine viral diarrhoea virus are all likely to negatively impact on calving rate. Recent work in Moorepark (J. Crilly, personal communication) indicates that herd size is a significant predictor of the incidence of salmonellosis in herds. Reproductive performance is often reduced following problems at or after calving, including dystocia, twins, retained foetal membranes and vaginal discharge (Morton, 2000). Lameness can be attributed to many factors including nutrition, genetics and cow facilities, and particularly cubicle numbers per cow. Concrete yards and cow pathways, where animals have to travel long distances to pasture, are often inadequate where originally designed for smaller herds. Lameness during early lactation can have substantial effects on reproductive performance, particularly among cows suffering severe lameness during the first 6 weeks of mating and these effects are also associated with decreased submission rates (Morton, 2000). The financial implications of lower calving rates and delays in conception have been detailed recently by Esslemont et al. (2001). They concluded that there is considerable extra profit available from relatively small improvements in fertility management, but that the concept of lost profit due to long calving intervals and high culling rates needs to be explained and promoted to farmers.
5. Conclusion It is well documented that reproductive performance has declined in dairy herds and that this decline has been in association with dramatic increases in milk production. Based on the results of this study, some of that temporal decline is likely to be due to increased herd size over the same period. Herd-level milk production was not associated with calving rate in the present study which may indicate that any adverse effects of high milk production on reproductive performance can be overcome by good management. Due mainly to economies of scale, large herd sizes are likely to be a significant feature of dairy farming systems in the future. Large herd sizes may contribute to reduced reproductive performance and may require new approaches to breeding management (Lucy, 2001). Data from the present study support this conclusion indicating that larger herds, in combination with other associated herd characteristics, are likely to have poorer calving rates than medium or small herds.
Acknowledgements The authors wish to acknowledge D. Cliffe for technical assistance and the farmers who participated in DairyMIS. References ACITS, 1995. Factor analysis using SAS Proc factor. Statistical services, University of Texas, Austin, USA. Available at: http://www.Utexas.edu/cc/docs/stat53.html. Accessed 6 April 2001. Buckley, F., Dillon, P., Mee, J.F., Veerkamp, R., 2000. Moorepark fertility project—latest results. In: Proceedings of Irish Grassland Association Dairy Conference. Tullamore, Ireland, pp. 24–37.
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
149
Chatfield, C., Collins, A.J., 1989. Introduction to Multivariate Analysis. Chapman & Hall, London. Crosse, S., 1985. Dairy Herd Management Information System—DairyMIS II. Part 2: Animal Health and Machine Milking. Moorepark 25th Anniversary Publication, pp. 325–361. Crowley, J.P., Harrington, D., Lacy, M., 1967. A survey of reproductive efficiency in cattle. Ir. J. Agric. Res. 6, 237–246. Cunningham, E.P., O’Byrne, T.M., Murphy, N., 1978. Survey of A.I. Results 1978: Final Report Trinity College Dublin. Applied Research and Consultancy Group, Dublin 2, Ireland. Cunningham, E.P., O’Byrne, T.M., 1980. Survey of A.I. Results 1980: Final Report Trinity College Dublin. Applied Research and Consultancy Group, Dublin 2, Ireland. Dillon, P., 1996. Maximising profit in creamery milk production using current research and technology. Ir. Grassland Anim. Prod. Assoc. J. 30, 8–21. Eneveloldsen, C., Sorensen, J.T., Thysen, I., Guard, C., Grohn, Y.T., 1995. A diagnostic and prognostic tool for epidemiologic and economic analyses of dairy herd health management. J. Dairy Sci. 78, 947–961. Eneveloldsen, C., Hindhede, J., Kristensen, T., 1996. Dairy herd management types assessed from indicators of health, reproduction, replacement and milk production. J. Dairy Sci. 79, 1221–1236. Esslemont, R.J., Kossaibiti, M.A., Allcock, J., 2001. Economics of fertility in dairy cows. In: Fertility in the High-Producing Dairy Cow, vol. 26. British Society of Animal Science Occasional Publication, pp. 19–29. Funk, D.A., 1993. Optimal genetic improvement for the high producing herd. J. Dairy Sci. 76, 3278–3286. Gilbert, R.P., Bailey, D.R.C., 1991. Hair coat characteristics and postweaning growth of Hereford and Angus cattle. J. Anim. Sci. 69, 498–506. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data Analysis. Prentice Hall, Upper Saddle River, New Jersey. Hoekstra, J., Van der Lugt, A.W., Van der Werf, J.H.J., Ouweltjes, W., 1994. Genetic and phenotypic parameters for milk production and fertility traits in upgraded dairy cattle. Livest. Prod. Sci. 40, 225–232. Holloway, J.W., Savell, J.W., Hamby, P.L., Baker, J.F., Stouffer, J.R., 1990. Relationships of empty-body composition and fat distribution to live animal and carcass measurements in Bos Indicus-Bos Taurus crossbred cows. J. Anim. Sci. 68, 1818–1826. Hurnik, D., Dohoo, I.R., Donald, A., Robinson, N.P., 1994. Factor analysis of swine farm management practices on Prince Edward Island. Prev. Vet. Med. 20, 135–146. Johnson, R.A., Wichern, D.W., 1982. Applied Multivariate Statistical Analysis. Prentice Hall, Englewood Cliffs, New Jersey. Lucy, M.C., 2001. Reproductive loss in high-producing dairy cattle: where will it end? J. Dairy Sci. 84, 1277–1293. Mee, J.F., Fahey, J., Crilly, J., 1999. Breeding the dairy cow of the future—todays challenges. In: Proceedings of the Teagasc National Dairy Conference on Dairying in the New Millennium. Adare, Teagasc, Moorepark, Fermoy, Co. Cork, Ireland, pp. 7–16. Morton, J.M., 2000. The in-calf project—some risk factors for reproductive performance in Australian dairy herds. In: Proceedings of the Australian and New Zealand Combined Dairy Veterinarians’ Conference. Le Lagon Resort, Port Vila, Vanuatu, Australia, ISSN 0112-9643, pp. 43–61. Nebel, R.L., McGilliard, M.L., 1993. Interactions of high milk yield and reproductive performance in dairy cows. J. Dairy Sci. 76, 3257–3268. O’Farrell, K.J., Crilly, J., 1998. Fertility rates in Irish dairy herds. In: Proceedings of the 20th World Buiatrics Congress, vol. 2. Sydney, Australia, pp. 585–587. Price, E.O., Wallach, S.J.R., 1991. Sexual and aggressive behaviours of Hereford bulls: interrelationship of variables and problems with the use of composite scores to describe overall sexual performance. J. Anim. Sci. 69, 1028–1033. Pryce, J.E., Esslemont, R.J., Thompson, R., Veerkamp, R.F., Kossaibati, M.A., Simm, G., 1998. Estimation of genetic parameters using health, fertility and production data from a management recording system for dairy cattle. Anim. Sci. 66, 577–584. Roche, J.F., Sherington, J., Mitchell, J.P., Cunningham, J.F., 1978. Factors affecting calving rate to A.I. in cows. Ir. J. Agric. Res. 17, 149–157. SAS User’s Guide: Statistics, 1989. Version 6, 4th ed., vol. 2. SAS Institute, Cary, NC. Shahin, K.A., Ashmawy, A.A., Mourad, K.A., Foda, T.A., 1998. Varimax rotated factors derived from body dimensions as used to predict production criteria in buffalo cows. Indian J. Dairy Sci. 68, 946–949.
150
J. Fahey et al. / Animal Reproduction Science 74 (2002) 133–150
Sieber, M., Freeman, A.E., Hinz, P.N., 1987. Factor analysis for evaluating relationships between first lactation type scores and production data of Holstein dairy cows. J. Dairy Sci. 70, 1018–1026. Stevenson, J.S., 2001. A review of oestrus behaviour and detection in dairy cows. In: Fertility in the High-Producing Dairy Cow, vol. 26. British Society of Animal Science Occasional Publication, pp. 43–62. Van Arendonk, J.A.M., Nieuwhof, G.J., Vos, H., Korver, S., 1991. Genetic aspects of feed intake and efficiency in lactating dairy heifers. Livest. Prod. Sci. 29, 263–275. White, D.S., O’Farrell, J.D., 1972. The artificial insemination service and conception rates. Vet. Services Bull., UK 1, 60–66.