Participants
The representative sample was calculated according to sex, for a finite population with a confidence level of 95% and a margin of error of +5%, 378 being girls (Mage=13.99; SD=.30) and 375 boys (Mage=14.02; SD=.33).
Item Descriptive Analysis
The item statistics for each of the dimensions are presented in Table 1. In the aggressive dimension, it should be noted that the SD of item 1 was <1 and the skewness and kurtosis values presented values <2. In addition to this, the internal consistency of the dimension with the four items was inadequate (α=.56) and does not improve if any of the items are eliminated. These results should be considered for the evaluation of model fit in confirmatory analyses. The items of the dimensions of low engagement or irresponsibility, fails to follow directions and distract or disturbs others, obtained adequate values in SD, CCIT-c, skewness, kurtosis, and reliability. The irresponsibility and low commitment factor obtained values slightly below those accepted in reliability; however, according to Taylor, Ntoumanis, and Standage [48], when a factor is composed of a small number of items (in this case by four items) an internal consistency index <.70 can be considered acceptable. Regarding the items of the poor self-management factor, the deletion of item 19 improved the internal consistency to .75 and item 20 obtained skewness and kurtosis values >2. These results should be considered for the evaluation of model fit in confirmatory analyses.
Table 1. Descriptive, internal consistency and homogeneity statistics (n = 315).
Dimensions:
|
M
|
SD
|
CCIT-c
|
α without item
|
Q1
|
Q2
|
Factor 1: Aggressive (α = .56)
|
|
|
|
|
|
|
1. Amenazo a los demás compañeros/as de clase. [Threatens others]
|
1.47
|
.95
|
.35
|
.48
|
2.23
|
4.39
|
2. Hablo correctamente a mis compañeros/as de clase. [Smart mouth toward students]
|
2.27
|
1.15
|
.38
|
.45
|
0.65
|
-0.32
|
3. Hablo de los demás a sus espaldas. [Talking back]
|
1.81
|
1.10
|
.32
|
.49
|
1.36
|
1.05
|
4. Hablo de forma correcta al profesor/a. [Smart mouth toward teacher]
|
1.90
|
1.17
|
.32
|
.50
|
1.34
|
0.91
|
Factor 2: Low Engagement or Irresponsibility (α = .69).
|
|
|
|
|
|
|
5. Me quejo habitualmente. [Whining]
|
2.57
|
1.33
|
.42
|
.61
|
0.41
|
-0.96
|
6. Soy perezoso en clase. [Lazy]
|
2.47
|
1.37
|
.42
|
.62
|
0.46
|
-1.01
|
7. Busco llamar la atención. [Attention seeking]
|
1.73
|
1.09
|
.45
|
.59
|
1.43
|
1.08
|
8. Me muevo lentamente a propósito. [Moves slowly on purpose]
|
1.89
|
1.22
|
.50
|
.56
|
1.23
|
0.37
|
Factor 3: Fails to Follow Directions (α = .71).
|
|
|
|
|
|
|
9. Interrumpo las clases. [Interrupts]
|
1.72
|
1.12
|
.53
|
.70
|
1.55
|
1.43
|
10. Me siento inseguro en clase de EF. [Unsafe actions]
|
1.96
|
1.29
|
.45
|
.74
|
1.10
|
-0.06
|
11. No presto atención en clase de EF. [Doesn’t pay attention]
|
2.07
|
1.37
|
.57
|
.67
|
1.02
|
-0.30
|
12. No sigo las instrucciones. [Not following directions]
|
1.86
|
1.17
|
.64
|
.64
|
1.23
|
0.41
|
Factor 4: Distract or Disturbs Others (α = .85).
|
|
|
|
|
|
|
13. Tengo mal genio y me enojo. [Temper tantrums]
|
2.08
|
1.26
|
.48
|
.83
|
0.92
|
-0.30
|
14. Abandono el grupo durante una actividad. [Leaving the group during an activity]
|
1.69
|
1.11
|
.69
|
.72
|
1.68
|
1.90
|
15. Miento en clase. [Lying]
|
1.71
|
1.11
|
.68
|
.73
|
1.58
|
1.54
|
16. Me salto las clases de EF. [Sneaks out of class]
|
1.51
|
1.00
|
.67
|
.74
|
2.10
|
3.62
|
Factor 5: Poor Self-Management (α = .70)
|
|
|
|
|
|
|
17. Soy peleónero/a. [Quarrelsome]
|
1.82
|
1.16
|
.58
|
.58
|
1.31
|
0.70
|
18. Me burlo de otros/as compañeros/as de clase. [Makes fun of other students]
|
1.89
|
1.18
|
.58
|
.57
|
1.22
|
0.45
|
19. Argumento mis actos. [Arguing]
|
2.76
|
1.39
|
.32
|
.75
|
0.17
|
-1.19
|
20. Acoso a algunos/as compañeros/as de clase. [Bullying]
|
1.49
|
1.08
|
.50
|
.63
|
2.29
|
4.12
|
Confimatory Factor Analysis and Reliability
Next, the original dimensionality theoretically proposed by Krech et al. [22] was analyzed with CFA and, following authors such as Markland [35], several models were formulated and analyzed, given that the data so recommended, and the most relevant results were reported. Considering the above in the analysis of scale items, it was appropriate to perform and compare several structural regression models to check the best fit. Several models were hypothesized (see Table 2). The first of the models included the 20 items and the five factors of the original scale [22] that presented some unacceptable fits (TLI=.87; CFI=.89), and two of the items of the aggressive dimension presented low regression weights <.33 [41] and high values (>2.58) in their standardized residuals [49]. Considering the above, as well as the unacceptable internal consistency of the aggressive factor (α=.56) this factor was eliminated and a model with the other four factors was evaluated, it presented acceptable goodness-of-fit indices. However, taking into account the index modification values of the statistical program, the errors of items 11 and 12 of the fails to follow directions factor and items 14 and 16 of the distract or disturbs others factor were correlated and the model presented acceptable (TLI=.94; CFI=.95) and excellent (SRMR=.036; RMSEA=.058) goodness-of-fit indices. Thus, H1 is fulfilled. Despite the good fits of this model, it was considered that the high correlations (.94) between some of the factors (i.e., distract or disturbs others with poor self-management and fails to follow directions with low poor self-management) could limit the discriminant validity of the scale, as values <.85 are considered adequate [50], although some authors consider values <. 90 to be [51], so a higher-order model [24] was evaluated, which also presented acceptable (TLI=.94; CFI=.95) and excellent (SRMR=.038; RMSEA=.060) goodness-of-fit indices. To examine the model comparison, the AIC (Akaike Information Criteria) and BIC (Bayesian Information Criterion) were also taken into account, in which, although they do not describe the model fit, lower values are considered to reflect a better fit.
Table 2 Fit indices for each model.
Models
|
χ2
|
df
|
p
|
χ2/gl
|
TLI
|
CFI
|
SRMR
|
RMSEA(90%IC)
|
AIC
|
BIC
|
Model 5 factors
|
765.80
|
160
|
.000
|
4.79
|
.87
|
.89
|
.049
|
.071(.066;.076)
|
865.80
|
1097.00
|
Models 4 factors
|
456.10
|
98
|
.000
|
4.65
|
.91
|
.92
|
.042
|
.070(.063;.076)
|
532.10
|
707.81
|
Models 4 factors*
|
339.49
|
96
|
.000
|
3.54
|
.94
|
.95
|
.036
|
.058(.051;.065)
|
419.49
|
604.45
|
Higher order model*
|
361.22
|
.98
|
.000
|
3.69
|
.94
|
.95
|
.038
|
.060(.053;.066)
|
437.22
|
612.94
|
Note. χ2 = chi-square; df = degrees of freedom; TLI = Tucker Lewis index; CFI = comparative fit index; SRMR = Standardized Root Mean-Square; RMSEA = root-mean squared approximation; IC = confidence interval; AIC = Akaike Information Criteria; BIC = Bayesian Information Criterion; *model with correlation of the errors of items 11 with 12 and 14 with 16.
In Table 3, we can see the standardized factor loadings for first-order CFA and higher order CFA of the CCD-EF with model four factors. To evaluate the reliability and validity of the scale, α, composite reliability and average variance extracted, and AVE were measured for each factor. The results can be seen in Table 4. The α values are acceptable and only the low engagement or irresponsibility dimension presented values <.70, although considering the stipulations of Taylor et al. [47], given the low number of items, this value can be accepted. All factors present acceptable composite reliability values [41]. In relation to the AVE, these same authors indicate that convergent validity values are considered acceptable when all the values of the standardized regression weights in a latent variable are significant and >.50, even if its AVE is <.50. This is the case for low engagement or irresponsibility and fails to follow directions. H2 is, thus, satisfied.
Table 3 Standardized factor loadings for first-order CFA and H-CFA of the CCD-EF.
|
CFA
|
H-CFA
|
Item
|
LEI
|
FFD
|
DDO
|
PSM
|
LEI
|
FFD
|
DDO
|
PSM
|
DB
|
Low engagement or irresponsibility (LEI)
|
|
|
|
|
|
|
|
|
.842**
|
Item5
|
.570**
|
|
|
|
.578**
|
|
|
|
|
Item6
|
.562**
|
|
|
|
.568**
|
|
|
|
|
Item7
|
.662**
|
|
|
|
.653**
|
|
|
|
|
Item8
|
.588**
|
|
|
|
.690**
|
|
|
|
|
Fails to follow directions (FFD)
|
|
|
|
|
|
|
|
|
.972**
|
Item9
|
|
.761**
|
|
|
|
.762**
|
|
|
|
Item10
|
|
.515**
|
|
|
|
.506**
|
|
|
|
Item11
|
|
.525**
|
|
|
|
.528**
|
|
|
|
Item12
|
|
.668**
|
|
|
|
.674**
|
|
|
|
Distract or disturbs others (DDO)
|
|
|
|
|
|
|
|
|
.974**
|
Item13
|
|
|
.577**
|
|
|
|
.576**
|
|
|
Item14
|
|
|
.745**
|
|
|
|
.748**
|
|
|
Item15
|
|
|
.794**
|
|
|
|
.794**
|
|
|
Item16
|
|
|
.743**
|
|
|
|
.743**
|
|
|
Poor self-management (PSM)
|
|
|
|
|
|
|
|
|
.944**
|
Item17
|
|
|
|
.720**
|
|
|
|
.722**
|
|
Item18
|
|
|
|
.696**
|
|
|
|
.697**
|
|
Item19
|
|
|
|
.354**
|
|
|
|
.353*
|
|
Item20
|
|
|
|
.713**
|
|
|
|
.711**
|
|
Correlations
|
|
|
|
|
|
|
|
|
|
FFD
|
.892**
|
|
|
|
|
|
|
|
|
DDO
|
.792**
|
.938**
|
|
|
|
|
|
|
|
PSM
|
.766**
|
.898**
|
.944**
|
|
|
|
|
|
|
Nota. CFA = first order confirmatory factorial analysis; H-CFA = higher order CFA; DB = disruptive behaviors; **p < 0.01.
Table 4. Reliability and validity of the CCD-EF factors.
Factors
|
Composite Reliability
|
AVE
|
Cronbach´s Alpha
|
Low engagement or irresponsibility
|
.72
|
.38
|
.69
|
Fails to follow directions
|
.71
|
.39
|
.75
|
Distract or disturbs others
|
.81
|
.52
|
.81
|
Poor self-management
|
.72
|
.41
|
.70
|
Note. AVE = average variance extracted.
Factorial Invariance Across Gender
The invariance of the CCD-EF was evaluated across gender (i.e., male=375, female=378) based on the first order CFA model and the higher order (H-CFA) model, the results of which are shown in Tables 5. Starting with a configural invariance model, invariance constraints were progressively added to the factor loadings (i.e., weak invariance, intercepts (i.e., strong invariance), and residual variances (i.e., strict invariance), weak invariance), intercepts (i.e., strong invariance), and residual variances (i.e., strict invariance). The values of these restrictive models were acceptable, except for the strict invariance of the H-CFA, as the CFI results were outside the cut-off values. The values of these restrictive models of the H-CFA did not exceed the cut-off points for RMSEA (Δ>.015), CFI (Δ>.01), and TLI (Δ>.01) so it can be considered fully invariant. In the case of the first-order model, it can be considered partially invariant, since the strict invariance model showed a decrease that slightly exceeded the limits of the recommended values (ΔCFI=-.013). H3 is, thus, fulfilled.
Table 5. Invariance test across gender for the CCD-EF.
Model
|
χ2
|
df
|
RMSEA [90% IC]
|
CFI
|
TLI
|
ΔRMSEA
|
ΔCFI
|
ΔTLI
|
Measurement across gender (First order CFA)
|
|
|
|
|
|
|
|
|
1.- Configural invariance
|
489.542*
|
192
|
.045 [.040;.050]
|
.936
|
.920
|
|
|
|
2.- Weak invariance
|
505.856*
|
204
|
.044 [.040;.049]
|
.935
|
.924
|
-.001
|
-.001
|
.004
|
3.- Strong invariance
|
553.927*
|
214
|
.046 [.041;.051]
|
.927
|
.918
|
.002
|
-.008
|
-.006
|
4.- Strict invariance
|
632.163*
|
232
|
.048 [.043;.052]
|
.914
|
.911
|
.002
|
-.013
|
-.007
|
Measurement across gender (H-CFA)
|
|
|
|
|
|
|
|
|
1.- Configural invariance
|
514.016*
|
196
|
.046 [.042;.051]
|
.932
|
.917
|
|
|
|
2.- Weak invariance
|
529.478*
|
208
|
.045 [.041;.050]
|
.931
|
.921
|
-.001
|
-.001
|
.004
|
3.- Strong invariance
|
554.529*
|
212
|
.046 [.042;.051]
|
.927
|
.917
|
.001
|
-.004
|
-.004
|
4.- Strict invariance
|
95.473*
|
212
|
.046 [.042;.051]
|
.927
|
.917
|
.000
|
.000
|
.000
|
Note. CFA = first order confirmatory factorial analysis; H-CFA = higher order CFA; χ 2= Chi square; df = degrees of freedom; RMSEA = root mean square error of approximation; 90%CI = 90% confidence interval of the RMSEA; CFI = comparative fit index; TLI = Tucker–Lewis index; * p < .01.
Nomological validity
A regression analysis with latent variables was performed to test the extent to which the dimensions of the SSI-EF (independent variable) predict disruptive behaviors (dependent variable). Firstly, to indicate that the SSI-EF showed excellent goodness-of-fit indices in CFA: χ2=34.30; gl=19; p=.01; χ2/gl=1.80; GFI=.99; CFI=.97, RMSEA=.04, SRMR=.05, as well as adequate reliability: satisfaction/fun, CR = .84; AVE=.52; α=78; boredom, CR=.71; AVE=.45; α=.65. Next, the measure of invariance of the linear regression model was tested according to the sex variable (girls, n=378; boys, n=375). The regression model was found to be invariant by sex, as the restrictive increase/decrease of the models for RMSEA was <.015, CFI<.01, and TLI<.01, the results are shown in Figure 1. In both boys and girls, the prediction of satisfaction/fun with PE on disruptive behaviors was not significant (H4 is not satisfied), but it was significant for boredom with PE (H5 is satisfied). When males are bored in PE classes, the predictive relationship for disruptive behaviors is .51 (p<.0001), with 23% of variance explained; while among female students the predictive relationship of boredom with PE for disruptive behaviors in class is .36 (p<.0001), with 15% of variance explained.