Personality and Individual Differences 112 (2017) 12–17
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Development of a short form of the Greek Big Five Questionnaire for Children (GBFQ-C-SF): Validation among preadolescents Angelos Markos, Constantinos M. Kokkinos ⁎ Department of Primary Education, School of Education Sciences, Democritus University of Thrace, Greece
a r t i c l e
i n f o
Article history: Received 10 December 2016 Received in revised form 15 February 2017 Accepted 19 February 2017 Available online xxxx Keywords: Big five Personality Preadolescents Assessment BFQ-C-Short Form
a b s t r a c t The Big Five Questionnaire for Children (BFQ-C) is a relatively new 65-item scale for assessing the Five Factor Model (i.e. Energy/Extraversion, Agreeableness, Conscientiousness, Emotional Instability, and Intellect/Openness) in childhood and early adolescence. The BFQ-C has been validated in several languages, including Italian, Dutch, German, Spanish and Greek. Nonetheless, its applicability may be limited by its length. This study derived an abbreviated form of the Greek translation of the BFQ-C, using techniques from both classical test theory and item response theory. The shortened form of the Greek BFQ-C (GBFQ-C-Short Form) included 30 items, with 6 items per subscale. The five-factor structure of the original BFQ-C was validated for a sample of Greek preadolescents and was used to develop the shortened version, which was then cross-validated using a second, independent sample. Results highlighted that the GBFQ-C-SF has equivalent psychometric properties compared with the full version and showed strong convergence with theoretically-relevant external criteria. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction The Five Factor Model (FFM) of personality, also known as the Big Five, is a useful descriptive taxonomy for most personality traits (John, Naumann, & Soto, 2008), and represents a powerful frame of reference in psychological reasoning about the structure of inter-individual differences in personality. The use of the FFM in the study of childhood personality has been the focus of recent research (e.g., Caspi & Shiner, 2006). There is compelling evidence that the FFM adequately represents the structure of personality in young children (e.g., Measelle, Ablow, John, Cowan, & Cowan, 2005), as well as in older children and adolescents (Soto, John, Gosling, & Potter, 2008; Tackett, Krueger, Iacono, & McGue, 2008). A common approach in the assessment of the five factors in children aged as young as 3 years old through adolescence is through parent (e.g., Halverson et al., 2003) and teacher reports (e.g., Mervielde, Buyst, & De Fruyt, 1995). However, a five-factor structure similar to the FFM has also been self-reported among 5th to 8th graders (Graziano, Jensen-Campbell, & Finch, 1997), as well as adolescents (e.g., McCrae & Costa, 2004). Successfully studying children and adolescent personality within the FFM framework necessitates psychometrically sound measures. Of the most common approaches in personality assessment among children and adolescents is the modification of item wording of well-known adult questionnaires to make them suitable for children. Within the ⁎ Corresponding author at: Department of Primary Education, Democritus University of Thrace, N. Hili, GR 68131, Alexandroupolis, Greece. E-mail address:
[email protected] (C.M. Kokkinos).
http://dx.doi.org/10.1016/j.paid.2017.02.045 0191-8869/© 2017 Elsevier Ltd. All rights reserved.
FFM, the use of this strategy was adopted in the construction of a children's version of the Big Five Questionnaire (BFQ-C; Barbaranelli, Caprara, Rabasca, & Pastorelli, 2003). The reliability and validity of the BFQ-C have been established for the Dutch (Muris, Meesters, & Diederen, 2005), German (Essau, Sasagawa, & Frick, 2006), Spanish (e.g., Holgado, Carrasco, del Barrio, & Chacόn, 2009) and Greek versions (Kokkinos & Markos, 2015). Although the BFQ-C presents strong psychometric properties and solid theoretical bases, its length (65 items) limits its applicability and may deter potential users. A shorter version of the BFQ-C may be more suitable for studies which include a battery of assessment measures and may be more likely to be completed in a valid manner by children and adolescents. A shortened form, however, does not automatically maintain the reliability and validity of the long form. Instead, it needs to be treated as a new instrument which should maintain the factor structure of the long form, has adequate content coverage of each original factor, provides adequate reliability and relates to external criteria in the same manner as the long form (Marsh, Ellis, Parada, Richards, & Heubeck, 2005; Myers, MacPherson, McCarthy, & Brown, 2003; Smith, McCarthy, & Anderson, 2000). To address these issues, this study aims to develop and assess a short version of the Greek BFQ-C. Previous attempts to create a short BFQ-C form were reported. For example, a short 30-item BFQ-C version has been used in a validation study of the Junior Version of the Spanish NEO-PI-R by Ortet et al. (2011). Item selection was based on high loadings on each intended factor and, at the same time, low loadings on the other four factors from the original 65-item BFQ-C. Another study by Zuffianò et al. (2013) used a
A. Markos, C.M. Kokkinos Personality and Individual Differences 112 (2017) 12–17
30-item reduced version of the BFQ-C but the authors did not provide any additional reference about the construction of this version. Moreover, Beatton and Frijters (2012) used a short-form Big Five Questionnaire for Children constructed by a selection of six questions per personality trait based on the psychometric properties reported in Muris et al. (2005). A limitation of all previous studies is that the authors did not employ any formal and rigorous scale development techniques. The present study reports on the development of a short 30-item version (6 items per factor) of the Greek BFQ-C (i.e. GBFQ-C-Short Form) and provides a more comprehensive evaluation of its construct validity. The 65-item BFQ-C has been recently validated on a large Greek preadolescent sample (GBFQ-C; Kokkinos & Markos, 2015). Based on data from this sample (Sample 1), both classical test theory (CTT) and item response theory (IRT) approaches were employed for item selection. IRT offers some advantages over classical methods for analyzing self-report personality-oriented data. IRT methods can model nonlinear relationships between the latent variable and observed scores, which rarely represent the empirical reality of behavioral constructs (Rusch, Lowry, Mair, & Treiblmaier, 2017). They can also provide information about the measurement process across the range of a latent trait at both the item and test level rather than providing only a single reliability estimate (Cooper & Petrides, 2010). This allows researchers to identify items that contribute most to measurement precision. Following item selection, to determine the generalizability of the shortened instrument beyond Sample 1, we examined the adequacy of its factor structure with an independent sample (Sample 2). Finally, to investigate the convergent validity of the GBFQ-C-SF, we explored associations between the shortened form and measures designed to tap related traits. Previous studies highlighted the relationship between Callous-Unemotional (CU) traits and personality (e.g., Essau et al., 2006; Roose, Bijttebier, Decoene, Claes, & Frick, 2010). Negative associations are generally expected between CU traits, Agreeableness and Conscientiousness. Further, Behavioral Activation System (BAS) sensitivity is expected to be positively associated with Extraversion (Smits & Boeck, 2006; Yu, Branje, Keijsers, & Meeus, 2011), whereas no association is expected between BAS and Emotional Instability (Yu et al., 2011).
2. Method 2.1. Participants and procedure The study was based on two independent preadolescent samples. The first included 1089 Greek elementary school students (10– 12 years old; M = 11.64; SD = 0.71) (males, n = 571; 51.8%). Two hundred eighty-six students were 10, 554 were 11, and 263 were 12 years old, who completed the original 65-item version of the GBFQ-C. Sample 2 included 347 students (11–16 years old; M = 13.05; SD = 0.70) (females, n = 193; 55.6%). One-hundred twenty-eight (128) of the participants were 11 to 12, and 219 were 13 to 16 years old, and completed a battery of self-report instruments including the GBFQ-C-SF, as well as measures of CU traits and BAS sensitivity (Section 2.2). Data screening revealed a small number of missing values in both datasets (items were missing between 0.0 and 3.2%, average 2.1%). School principals were conducted by the authors who explained the purpose and the procedure of the study. Then, teachers were asked to identify those students with no reading difficulties and to whom parental consent forms were handed out, and students were asked to return them only if their parents did not wish them to participate in the project. The voluntary basis of participation was also stressed. Less than ten students did not agree to participate.
13
2.2. Measures 2.2.1. Personality The Greek Big Five Questionnaire for Children (Barbaranelli et al., 2003; Kokkinos & Markos, 2015) is a 65-item scale assessing the five basic factors of personality in children, as young as 8 years, and adolescents; each factor consists of 13 items. Energy/Extraversion (E/E) assesses characteristics such as activity, enthusiasm, assertiveness and self-confidence. Agreeableness (A) taps concern and sensitivity toward others. Conscientiousness (C) assesses dependability, precision, and commitment fulfillment. Emotional Instability (EI) refers to feelings of anxiety, depression, and anger. Finally, Intellect/Openness (I/O) taps intellect, broadness or narrowness of cultural interests, and creativity. Responses are given on a 5-point scale ranging from “almost never” to “almost always” maintaining the original scoring system. 2.2.2. Callous-unemotional traits The 24-item self-report Inventory of Callous-Unemotional Traits (ICU; Frick, 2004) was used. Confirmatory Factor Analysis (CFA; Kokkinos & Voulgaridou, 2016a) provided evidence for a 3-factor model for CU traits: Callousness (e.g., “the feelings of others are unimportant to me”), Unemotional (e.g., “I hide my feelings from others”), and Uncaring (e.g., “I try not to hurt others' feelings”). The ICU is composed of 12 positively and 12 negatively worded items rated on a 4point scale (0 = “not at all true” to 3 = “definitely true”). Its validity was supported in previous research (Roose et al., 2010). Internal consistency of callous, unemotional and uncaring scales in Sample 2 was satisfactory (Cronbach's alpha 0.78, 0.68, and 0.80 respectively). 2.2.3. BAS sensitivity Behavioral activation system sensitivity was assessed with the 13item self-report version of Carver and White's (1994) BAS adaptation for children (Kokkinos & Voulgaridou, 2016b; Muris et al., 2005). The scale assesses BAS sensitivity at a cognitive level, and therefore measures the consequences of an activity and not the activity of the system per se (Smits & Boeck, 2006). BAS sensitivity is reflected in three empirically derived scales: Reward Responsiveness (5 items, e.g., “When I get something I want, I feel excited and energized”), Drive (4 items, e.g., “When I want something, I usually go all out to get it”), and Fun Seeking (4 items, e.g., “I crave excitement and new sensations”). Item responses ranged from 1 (“Strongly disagree”) to 4 (“Strongly agree”). Total scores were computed by summing item responses for each subscale and the scores of the three subscales were summed to obtain a total BAS score. Reliability was moderate to high in Sample 2, with alphas of 0.77 (Total), 0.68 (Reward Responsiveness), 0.72 (Drive), and 0.71 (Fun Seeking). 2.3. Data analysis Item selection was performed combining methods derived from CTT and IRT. The 13 original items of each subscale were first subjected in separate IRT analyses, which were conducted using Samejima's Graded Response Model (Samejima, 1969) in Mplus 7.3 (Muthén & Muthén, 2014). The IRT unidimensionality assumption was assessed for each subscale using exploratory factor analysis (EFA) with principal axis factoring on polychoric correlation matrices. The number of factors extracted was based on the ratio of the first to the second eigenvalue, which should be above 3 for unidimensionality to be considered appropriate (Morizot, Ainsworth, & Reise, 2007). Local independence was evaluated by examining the residual correlations between items. Items were assumed to be locally independent if the absolute value of the residual correlation between two items was 0.20 or less. Item information curves were inspected over a wide range of ability levels, Θ, ranging from − 6.0 to 6.0, to select items for inclusion in the short form. Pairs of items with similar information curves (i.e. similar difficulty and discrimination) provide redundant information and where this
14
A. Markos, C.M. Kokkinos Personality and Individual Differences 112 (2017) 12–17
Table 1 Percentage of psychometric information retained by the GBFQ-C-Short Form scales. Subscale
Best items retained
% info total
% increase average item info
r
Conscientiousness Intellect/Openness Agreeableness Emotional Instability Energy/Extraversion Average
22, 25, 28, 34, 44, 53 12, 18, 24, 30, 46, 62 11, 27, 32, 38, 45, 51 4, 6, 8, 15, 17, 49 1, 19, 23, 26, 40, 57
60 72 63 59 65 64
33 56 31 29 38 37
0.92 0.84 0.87 0.88 0.88 0.88
Note. *p b 0.05. **p b 0.01.
applied, elimination of one of the dependent items was considered. The best six items of each subscale were selected based on the amount of psychometric (Fisher) information (see e.g., Embretson & Reise, 2000) provided across a wide range of ability, Θ by the short form relative to the long form. Then, in the context of CTT, a CFA model was estimated using the long form. This model included five correlated factors representing the GBFQ-C subscales. From this model, in line with recommendations by Marsh et al. (2005), Myers et al. (2003) and Smith et al. (2000), the best six items from each factor were selected using the following criteria: (a) items with high standardized factor loadings and communalities, low correlated errors within factors and minimal cross-loadings on other factors, as indicated by model modification indices, (b) items that exhibit a wide coverage of the construct's content and (c) retain sufficient items in each factor such that reliabilities for each factor reflect those of the equivalent long form. Last, the results obtained from IRT and CTT analyses were compared and combined, and the final sets of items to be included in the 30-item short version were formed. After item selection, the adequacy of the measurement model underlying the short form which preserves the factor structure of the long form was verified using CFA on the data from the Sample 1, and validated using the data from the independent Sample 2. The fit of CFA models was obtained using WLSMV (weighted least-squares means and variance adjusted) estimation in Mplus 7.3. Model fit was evaluated using the chi-square (χ2) test, the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the Root Mean Square Error of Approximation (RMSEA) and its 90% confidence interval. Values N0.90 and 0.95 for both the CFI and TLI are considered to indicate adequate and excellent fit to the data, respectively, while values smaller than 0.08 or 0.06 for the RMSEA reflect acceptable and excellent model fit (Hu & Bentler, 1999). Internal consistency reliability estimates were calculated for each subscale using Cronbach's alpha and McDonald's omega coefficients. Finally, the convergent validity of the GBFQ-C-SF was examined through its correlation with theoretically-relevant external criteria. 3. Results 3.1. Item selection The suitability of implementing unidimensional IRT models for each bigfive subscale was first investigated. Since there were no item-to-item residual correlations with an absolute value larger than 0.20 the local independence assumption was fulfilled. Examination of the corresponding scree plots suggested that a majority of the variance was accounted for by a single factor and the ratios of first to the second eigenvalue were all above 3 in all cases. In sum, the EFA analyses suggested that there was one sufficiently dominant factor present in the data for each subscale, justifying the use of a unidimensional IRT model. The results of IRT analyses used for item selection are summarized in Table 1. The table reports the best six items of each original subscale that maintain the largest percentage of total psychometric information across the range of ability, Θ (−6.0 to 6.0). These percentages, shown in the third column, range from a low of 59% (Emotional Instability) to a high of 72% (Openness). On average, these 6-item short form scales captured 64% of the total psychometric information
of the 13-item subscales on the full length GBFQ-C. The percentage increase in the average item information for the 6 items relative to the original 13 items is shown in the fourth column of Table 1. For example, the average psychometric information across the range of ability captured by the best six items of Openness is 4.01, compared to a 2.57 for the 13 items of the long form. This corresponds to a 56% increase. Across all subscales, the information provided by each item was on average 37% higher for the 6-item subscales than for the 13-item subscales. Furthermore, a visual inspection of item characteristic curves revealed that none of the selected items in each shortened subscale provided redundant information. The convergent correlations between the long and short versions of each subscale (last column of Table 1) were high, ranging from 0.84 for Intellect/Openness to 0.92 for Conscientiousness, with an average of 0.88 across the five subscales. The item parameters of slope (discrimination) and thresholds (difficulty) of the IRT model are available upon request by the corresponding author. In the context of CTT, the hypothesized five-factor CFA model had acceptable fit implying adequate construct validity of the 65-item (full) version, χ2(2005) = 4255.71, p b 0.001, CFI = 0.95, TLI = 0.94, RMSEA =0.039, 90%CI = 0.038–0.041. The parameter estimates from this model are reported in Table 2 (factor loadings, uniquenesses, and reliability estimates). Standardized loadings were relatively high, ranging in absolute value, from 0.35 to 0.73 and statistically significant (ps b 0.01), with a satisfactory level of composite reliability (McDonald's ω = 0.79 to 0.83; Mean ω = 0.81). Correlations among latent factors ranged from 0.03 for Emotional Instability and Energy/Extraversion to a surprisingly high of 0.76 for Conscientiousness and Intellect/Openness (see a discussion of the correlation patterns in Kokkinos & Markos, 2015). All items retained previously via the IRT analyses demonstrated high standardized loadings on the respective factors (≥0.60), relatively low uniquenesses (b0.65) and low modification indices (b 4.0). This indicates that there were no significant cross-loadings and correlated errors. Therefore, IRT analyses were found to generally agree with those of CTT. Note that the results of CFA justified the inclusion of alternative items in the short version. An examination of the corresponding information curves, however, did not provide adequate support for inclusion. Table 3 presents the final sets of items (from the original BFQ-C) included in each abbreviated subscale.1 3.2. Adequacy of the factor structure of the short form Next, a CFA on the 30-item BFQ-C, specifying a measurement model with all items loading on their intended factor, provided an excellent fit to the data of the first sample, χ2(395) = 986.34, p b 0.001, CFI = 0.98, TLI = 0.97, RMSEA = 0.042, with a 90% CI = 0.039–0.045. The parameter estimates from this model are shown in Table 2 and inter-correlations among the five latent factors of the long and short form are presented in Table 4. Results indicated that all factors from the short scale present a satisfactory level of composite reliability (ω = 0.80 to 0.83; Mean ω = 0.82), and are well defined by moderate to high factor loadings ranging from 0.62 to 0.75, with a mean loading of 0.67. It is important to highlight that factor loadings, latent factor inter-correlations 1 The English translation of the BFQ-C-Short Form and the original Greek version (GBFQ-C-SF) are available upon request from the corresponding author.
A. Markos, C.M. Kokkinos Personality and Individual Differences 112 (2017) 12–17
15
Table 2 Parameter estimates for the CFA measurement models. Items
Sample 1, long version
Sample 1, short version
Loading
Uniqueness
Conscientiousness C3 C7 C20 C22S C25S C28S C34S C37 C44S C48 C53S C56 C65 Reliability (ω) Reliability (α)
0.54 0.51 0.50 0.67 0.67 0.67 0.69 0.48 0.65 0.53 0.63 0.40 0.66 0.83 0.81
0.71 0.74 0.75 0.55 0.55 0.55 0.52 0.77 0.58 0.72 0.60 0.84 0.56
Intellect/Openness O5 O10 O12S O18S O24S O30S O33 O36 O43 O46S O52 O59 O62S Reliability (ω) Reliability (α)
0.43 0.38 0.65 0.69 0.61 0.65 0.37 0.39 0.50 0.70 0.69 0.52 0.63 0.80 0.76
0.82 0.86 0.58 0.52 0.63 0.58 0.86 0.85 0.75 0.51 0.47 0.73 0.60
Agreeableness A2 A11S A13 A16 A21 A27S A32S A38S A45S A47 A51S A60 A64 Reliability (ω) Reliability (α)
0.38 0.66 0.45 0.47 0.45 0.69 0.67 0.71 0.61 0.48 0.60 0.35 0.40 0.83 0.81
0.86 0.56 0.80 0.78 0.80 0.52 0.55 0.50 0.63 0.77 0.64 0.88 0.84
Emotional Ιnstability N4S N6S N8S N15S N17S N29 N31 N39 N41 N49S N54 N58 N61 Reliability (ω) Reliability (α)
0.64 0.65 0.68 0.71 0.73 0.52 0.48 0.35 0.46 0.71 0.50 0.44 0.52 0.83 0.82
0.59 0.58 0.54 0.50 0.47 0.73 0.77 0.88 0.79 0.50 0.75 0.81 0.73
Energy/Extraversion E1S E9 E14 E19S
0.60 0.41 0.57 0.63
0.64 0.83 0.67 0.60
Sample 2, short version
Loading
Uniqueness
Loading
Uniqueness
0.71 0.68 0.70 0.69
0.50 0.54 0.51 0.52
0.79 0.66 0.71 0.73
0.38 0.56 0.50 0.47
0.66
0.56
0.69
0.52
0.64
0.59
0.68
0.54
0.83 0.82
0.82 0.79
0.70 0.70 0.63 0.67
0.51 0.51 0.60 0.56
0.76 0.67 0.63 0.73
0.42 0.55 0.60 0.47
0.73
0.47
0.65
0.58
0.65 0.80 0.77
0.58
0.68 0.79 0.76
0.54
0.66
0.56
0.64
0.59
0.70 0.69 0.73 0.64
0.50 0.52 0.47 0.59
0.62 0.69 0.73 0.64
0.62 0.52 0.47 0.59
0.62
0.62
0.61
0.63
0.83 0.81
0.84 0.70
0.65 0.66 0.72 0.70 0.75
0.58 0.56 0.48 0.50 0.44
0.65 0.67 0.79 0.73 0.74
0.58 0.55 0.38 0.47 0.45
0.70
0.51
0.62
0.62
0.83 0.80
0.83 0.76
0.62
0.62
0.74
0.45
0.68
0.54
0.80
0.36 (continued on next page)
16
A. Markos, C.M. Kokkinos Personality and Individual Differences 112 (2017) 12–17
Table 2 (continued) Items
Sample 1, long version
Sample 1, short version
Sample 2, short version
Loading
Uniqueness
Loading
Uniqueness
Loading
Uniqueness
E23S E26S E35 E40S E42 E50 E55 E57S E63 Reliability (ω) Reliability (α)
0.66 0.60 0.37 0.63 0.33 0.33 0.35 0.63 0.66 0.79 0.75
0.58 0.64 0.86 0.60 0.89 0.89 0.87 0.60 0.56
0.65 0.62
0.58 0.62
0.61 0.65
0.63 0.58
0.63
0.60
0.80
0.36
0.64
0.59
0.66
0.56
0.80 0.76
0.79 0.73
Note: Items retained in the short form (bold); α = Cronbach's alpha; ω = McDonald's omega composite reliability coefficient; All factors loadings are significant at p b 0.01.
and reliability estimates are of similar magnitude across the short and long forms. The psychometric properties of the GBFQ-C-SF were then validated using data from the second, independent sample (n = 347). The results from this model fully replicated those from Sample 1, showing an excellent fit to the data, (χ2(395) = 899.43, p b 0.001, CFI = 0.98, TLI = 0.97, RMSEA = 0.042, with a 90% CI = 0.039–0.045), moderate to high factor loadings ranging from 0.61 to 0.80, with a mean loading of 0.69, and high composite reliability (ω = 0.79 to 0.83; Mean ω = 0.81). The pattern of latent factor inter-correlations remained in line with those observed for the long form. 3.3. Convergent validity Correlations with callous-unemotional traits and BAS sensitivity for the five personality factors of the GBFQ-C-SF are shown in Table 5. First, Conscientiousness and Agreeableness were negatively and significantly correlated with all CU traits. The Uncaring subscale showed the strongest associations with the Big Five dimensions. Moreover, negative Table 3 The GBFQ-C-Short Form scales and items. Conscientiousness
22:During class-time I am concentrated on the things I do 25:When I finish my homework, I check it many times to see if I did it correctly 28:I respect the rules and the order 34:If I take an engagement I keep it 44:When I start to do something I have to finish it at all costs 53:I play only when I finished my homework Intellect/Openness 12:I easily learn what I study at school 18:When the teacher asks questions I am able to answer correctly 24:I like to read books 30:When the teacher explains something I understand immediately 46:I am able to solve mathematics problems 62:I understand immediately Agreeableness 11:I behave correctly and honestly with others 27:I treat my peers with affection 32:I behave with others with great kindness 38:I am polite when I talk with others 45:If a classmate has some difficulty I help her/him 51:I treat kindly also persons who I dislike Emotional 4:I get nervous for silly things Ιnstability 6:I am in a bad mood 8:I argue with others with excitement 15:I easily get angry 17:I quarrel with others 49:I easily lose my calm Energy/Extraversion 1:I like to meet with other people 19:I like to be with others 23:I can easily say to others what I think 26:I say what I think 40: I like to talk with others 57:I easily make friends
correlations were observed between the Uncaring dimension and Openness (r = −0.26, p b 0.01) and the Unemotional dimension and Extraversion (r = − 0.32, p b 0.01). As expected, BAS sensitivity dimensions were positively and significantly associated mostly with Extraversion. Finally, consistent with our expectations, Emotional Instability was not significantly correlated with any of the BAS dimensions. 4. Discussion The BFQ-C is a relatively new and promising self-report instrument for measuring the FFM in children as young as 8 years old. Existing research has provided support for the psychometric properties of the BFQ-C in different populations. Nonetheless, its overall length (65 items) may constrain its utility in some settings. Thus, the purpose of the present study was to develop a short form of each BFQ-C subscale and evaluate the extent to which it assessed the same content as the full measure among Greek preadolescents. Using a combination of IRT and CTT approaches, we initially identified the six items from each GBFQ-C subscale that provided the greatest amount of psychometric information about the latent trait. Results indicated that most of the variance in the longer version of each subscale is retained by the short form. CFA on the retained items supported the adequacy of the five-factor model. Finally, the GBFQ-C-SF factor structure was replicated and confirmed on a new, independent sample. A critical component of the present study was to assess the convergence of the short scale with theoretically-relevant external criteria. In this regard, results provided strong support for the ability of the GBFQ-C-SF factors to replicate the validity of the original scale, resulting in theoretically expected correlations with CU traits and the BAS sensitivity (Essau et al., 2006; Roose et al., 2010). There are some limitations to the current study. One potential concern is that the BFQ-C is a relatively new measure and has not, therefore, been widely adopted. More studies are needed to provide support for the validity and utility of both the full and short forms of the BFQ-C. A second limitation is that all data were collected via self-report questionnaires. Although there is considerable empirical support for the validity
Table 4 Intercorrelations among GBFQ-C and GBFQ-C-SF latent factors (Sample 1).⁎ C Conscientiousness (C) Intellect/Οpenness (I/O) Agreeableness (A) Emotional Ιnstability (EI) Energy/Εxtraversion (E/E)
0.76⁎⁎ 0.68⁎⁎ −0.34⁎⁎ 0.41⁎⁎
I/O
A
EI
E/E
0.75⁎⁎
0.59⁎⁎ 0.55⁎⁎
−0.31⁎⁎ −0.25⁎⁎ −0.29⁎⁎
0.31⁎⁎ 0.56⁎⁎ 0.49⁎⁎
0.62⁎⁎ −0.21⁎⁎ 0.68⁎⁎
−0.22⁎⁎ 0.59⁎⁎
−0.08 0.03
Note. Factor correlations for the long version under the diagonal. Factor correlations for the short version above the diagonal. ⁎⁎ p b 0.01. ⁎ p b 0.05.
A. Markos, C.M. Kokkinos Personality and Individual Differences 112 (2017) 12–17 Table 5 Pearson correlations with external criteria for the short version (Sample 2). Scale
C
I/O
A
EI
E/E
BAS Drive BAS Fun Seeking BAS Reward Responsiveness Total BAS ICU Callousness ICU Uncaring ICU Unemotional
−0.05 −0.09 0.07 −0.06 −0.30⁎⁎ −0.56⁎⁎ −0.20⁎⁎
−0.09 −0.02 0.10 −0.03 −0.08 −0.24⁎⁎ −0.04
−0.04 0.01 0.16⁎ 0.04 −0.35⁎⁎ −0.60⁎⁎ −0.34⁎⁎
−0.04 0.06 0.03 0.05 0.05 0.09 −0.10
0.16⁎⁎ 0.15⁎⁎ 0.25⁎⁎ 0.24⁎⁎ −0.06 −0.08 −0.32⁎⁎
Note. C = Conscientiousness, I/O = Intellect/Openness, A = Agreeableness, EI = Emotional Instability, E/E = Energy/Extraversion. ⁎⁎ p b 0.01. ⁎ p b 0.05.
of self-report measures of personality (e.g., Widiger & Boyd, 2009), selfreport data can be vulnerable to social desirability and response bias. Future studies should also include interviews and parent reports of personality. Third, because the GBFQ-C and GBFQ-C-SF have only been validated in Greek, future research should focus on establishing their cross-linguistic validity. Last, the reported convergent validity correlation coefficients between the shortened form and related traits, provide only weak evidence for the validity of the short form. Stronger evidence would be given, if the corresponding correlations with these variables were obtained for both the long and the short form. However, there are no data available from Study 1 for investigating these relationships. Despite these limitations, the GBFQ-C-SF is promising in terms of research utility. Overall, while previous attempts to create a short form of the BFQ-C were all based on traditional, sample-dependent approaches, this study relied on a combination of CTT and IRT techniques in a more systematic effort. The GBFQ-C-SF demonstrates sound psychometric properties and provides an assessment of childhood and early adolescence personality that is comparable to that achieved using the longer, 65-item GBFQ-C version. This suggests that the GBFQ-C-SF balances brevity with strong psychometric properties, it is more practical to administer and likely to be more appealing to a variety of researchers who might benefit from using this tool. References Barbaranelli, C., Caprara, G. V., Rabasca, A., & Pastorelli, C. (2003). A questionnaire for measuring the big five in late childhood. Personality and Individual Differences, 34, 645–664. Beatton, T., & Frijters, P. (2012). Unhappy Young Australian: A domain approach to explain life satisfaction change in children. QUT School of Economics and Finance Working/discussion paper (pp. 289). Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS scales. Journal of Personality and Social Psychology, 67, 319–333. Caspi, A., & Shiner, R. L. (2006). Personality development. In W. R. M. Lerner, N. Eisenberg, & W. Damon (Eds.), Handbook of child psychology. Vol. 3. Social, emotional and personality development (pp. 300–365) (6th ed.). New York, NY: Wiley. Cooper, A., & Petrides, K. V. (2010). A psychometric analysis of the trait emotional intelligence questionnaire–short form (TEIQue–SF) using item response theory. Journal of Personality Assessment, 92(5), 449–457. Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah, NJ: Lawrence Erlbaum Associates Publishers. Essau, C. A., Sasagawa, S., & Frick, P. J. (2006). Callous-unemotional traits in a community sample of adolescents. Assessment, 13, 454–469. Graziano, W. G., Jensen-Campbell, L. A., & Finch, J. F. (1997). The self as a mediator between personality and adjustment. Journal of Personality and Social Psychology, 73, 392–404. Halverson, C. F., Havill, V. L., Deal, J., Baker, S. R., Victor, J. B., Pavlopoulos, V., ... Wen, L. (2003). Personality structure as derived from parental ratings of free descriptions of children: The inventory of child individual differences. Journal of Personality, 71, 995–1026.
17
Holgado, F. P., Carrasco, M. A., del Barrio, M. V., & Chacόn, S. (2009). Factor analysis of the big five questionnaire using polychoric correlations in children. Quality and Quantity, 43, 75–85. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55. John, O. P., Naumann, L. P., & Soto, C. J. (2008). Paradigm shift to the integrative big five trait taxonomy: History, measurement, and conceptual issues. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of personality: Theory and research (pp. 114–158). New York, NY: Guilford Press. Kokkinos, C. M., & Markos, A. (2015). The big five questionnaire for children (BFQ-C): Factorial invariance across gender and age in a Greek sample of pre-adolescents. European Journal of Psychological Assessment (Advance online publication). Kokkinos, C. M., & Voulgaridou, I. (2016a). Relational victimization, callous-unemotional traits and hostile attribution bias among pre-adolescents. Journal of School Violence (Advance online publication). Kokkinos, C. M., & Voulgaridou, I. (2016b). Links between relational aggression, parenting and personality among adolescents. The European Journal of Developmental Psychology (Advance online publication). Marsh, H. W., Ellis, L. A., Parada, R. H., Richards, G., & Heubeck, B. G. (2005). A short version of the self description questionnaire-II: Operationalizing criteria for short-form evaluation with new applications of confirmatory factor analyses. Psychological Assessment, 17, 81–102. McCrae, R. R., & Costa, P. T., Jr. (2004). A contemplated revision of the NEO five- factor inventory. Personality and Individual Differences, 36, 587–596. Measelle, J. R., Ablow, J. C., John, O. P., Cowan, P. A., & Cowan, C. P. (2005). Can children provide coherent, stable, and valid self-reports on the big five dimensions? A longitudinal study from ages 5 to 7. Journal of Personality and Social Psychology, 89, 90–106. Mervielde, I., Buyst, V., & De Fruyt, F. (1995). The validity of the big-five as a model for teachers' ratings of individual differences among children aged 4–12 years. Personality and Individual Differences, 18, 525–534. Morizot, J., Ainsworth, A. T., & Reise, S. P. (2007). Toward modern psychometrics: Application of item response theory models in personality research. In R. W. Robins, R. C. Fraley, & R. F. Krueger (Eds.), Handbook of Research Methods in Personality Psychology (pp. 407–423). Guilford Press. Muris, P., Meesters, C., & Diederen, R. (2005). Psychometric properties of the big five questionnaire for children (BFQ-C) in a Dutch sample of young adolescents. Personality and Individual Differences, 38, 1757–1769. Muthén, L. K., & Muthén, B. O. (1998–2014). Mplus user's guide. Los Angeles, CA: Muthén & Muthén. Myers, M. G., MacPherson, L., McCarthy, D. M., & Brown, S. A. (2003). Constructing a short form of the smoking consequences questionnaire with adolescents and young adults. Psychological Assessment, 15, 163–172. Ortet, G., Ibáñez, M. I., Moya, J., Villa, H., Viruela, A., & Mezquita, L. (2011). Assessing the five factors of personality in adolescents: The junior version of the Spanish NEO-PI-R. Assessment, 19, 114–130. Roose, A., Bijttebier, P., Decoene, S., Claes, L., & Frick, P. J. (2010). Assessing the affective features of psychopathy in adolescence: A further validation of the inventory of callous and unemotional traits. Assessment, 17, 44–57. Rusch, T., Lowry, P. B., Mair, P., & Treiblmaier, H. (2017). Breaking free from the limitations of classical test theory: Developing and measuring information systems scales using item response theory. Information Management, 54, 189–203. Samejima, F. (1969). Estimation of latent ability using a response pattern of graded scores. Psychometrika monograph supplement. 34. (pp. 100–114). Smith, G. T., McCarthy, D. M., & Anderson, K. G. (2000). On the sins of short-form development. Psychological Assessment, 12, 102–111. Smits, D. J. M., & Boeck, P. D. (2006). From BIS/BAS to the big five. European Journal of Personality, 20, 255–270. Soto, C. J., John, O. P., Gosling, S. D., & Potter, J. (2008). The developmental psychometrics of big-five self-reports: Acquiescence, factor structure, coherence, and differentiation from ages 10 to 20. Journal of Personality and Social Psychology, 94, 718–737. Tackett, J. L., Krueger, R. F., Iacono, W. G., & McGue, M. (2008). Personality in middle childhood: A hierarchical structure and longitudinal connections with personality in late adolescence. Journal of Research in Personality, 42, 1456–1462. Widiger, T. A., & Boyd, S. (2009). Personality disorders assessment instruments. In J. N. Butcher (Ed.), Oxford handbook of personality assessment (pp. 336–363). New York: Oxford University Press. Yu, R., Branje, S. J., Keijsers, L., & Meeus, W. H. (2011). Psychometric characteristics of carver and White's BIS/BAS scales in Dutch adolescents and their mothers. Journal of Personality Assessment, 93, 500–507. Zuffianò, A., Alessandri, G., Gerbino, M., Kanacri, B. P. L., Di Giunta, L., Milioni, M., & Caprara, G. V. (2013). Academic achievement: The unique contribution of self-efficacy beliefs in self-regulated learning beyond intelligence, personality traits, and selfesteem. Learning and Individual Differences, 23, 158–162.