2. El proceso de formación de Ley en El Salvador
2.4 Etapas del proceso de formación de ley
2.4.4 Sanción
Although the DRAOR was constructed theoretically with three scales, at the time this study was conducted, there were no empirical investigations of this structure. In order to examine empirically the existing structure of the DRAOR, a Confirmatory Factor Analysis (CFA) was performed using version 19 of SPSS AMOS for Windows. Levels of fit were assessed by inspecting the significance levels of the chi-square, the comparative fit index (CFI), and the root mean square error of approximation (RMSEA). A non-significant chi-square is preferable; it indicates that the model is consistent with the observed data (Marsh & Hocevar, 1985). However, there are a number of problems with using significance levels for the chi-square as a guide to fit. The chi-square is likely to produce conservative estimates of fit when the sample is large, and when there are many variables and degrees of freedom (i.e., it produces a significant result even when there is a reasonably good fit to the data). The CFI fit index, unlike the chi-square test, is not affected by sample size and is considered to be a better measure of fit. A CFI value of .90 or higher is usually considered to indicate a good fit (Bentler, 1995). The RMSEA provides a measure of discrepancy per degree of freedom. A RMSEA value of .08 or lower is generally considered to reflect a reasonable fit (Browne & Cudek, 1993).
The three-factor model showed a reasonably poor fit to the data, 2(19) =
394.21, p < .001 (CFI = .84, RMSEA = .076). The chi-square was significant and the CFI was well below .90; the RMSEA on the other hand suggested a reasonable fit. Because of the poor fit of the existing three DRAOR subscales, I next conducted a series of Principal Components Analyses (PCA).
The dataset met the initial requirements for PCA. The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO = .86, exceeding the recommended value of .6 (Kaiser, 1970, 1974). Bartlett’s test of sphericity 2(153) =
1523.27, p < .001, indicated that correlations between items were sufficiently large. An initial unrotated solution yielded four components with eigenvalues over Kaiser’s
criterion of 1, explaining 29.8%, 8.6%, 7.6%, and 7.0% of the variance respectively. An inspection of the scree plot revealed a clear break after one and four components. Parallel analysis suggested retaining two components. Because of the disagreement in the number of components to retain from these methods, one-, two-, three- and four- component solutions were forced and compared on their interpretability, using orthogonal rotation.33
The four-component solution was chosen as the best and most interpretable fit for the data. This solution explained 51.5% of the variance. The eigenvalues for the unrotated solution were 5.52, 1.62, 1.38 and 1.27. In the rotated solution, the
eigenvalues for the four components evened out to 3.40, 2.90, 1.84 and 1.65. Loadings for each component are set out in Table 8 (loadings for the other solutions can be found in Appendix E). It was chosen over other solutions because in the rotated solution all components showed a number of strong loadings and each variable loaded
substantially on a component.34 There were no low communalities, suggesting that all
items fitted well with the other items in their component. Each component had at least three items and no items were omitted. Finally, the contents of each component were the most coherent theoretically.
Component 1 (Protective) consisted of all six protective items, making it identical to the protective subscale in Serin’s (2007) postulated three-factor structure. The strongest loadings were for “pro-social identity”, “high expectations”, and “social control”. Component 2 (Stable) comprised five stable items and one acute item. The strongest loadings were for the stable items “problem solving” and “entitlement”. Component 3 (labelled Internal Acute) consisted of three acute items: “negative mood”, “anger/hostility”, and “substance abuse”. Component 4 (labelled External Acute) consisted of four items: “interpersonal relationships”, “employment”, and “living
33 Oblique (oblimin) rotation was first performed due to expected correlations between the components. The solutions were the same as for the orthogonal rotation; therefore only the orthogonal solution was presented here for ease of interpretability.
34 Two items (“living situation” and “employment”) had loadings of .4 or above on two different components; these items were assigned to the component for which the loading was highest.
situation” (acute items), and “attachment with others” (stable item). Pearson’s correlations between the components are presented in Table 9; they are all in the expected directions.
Table 8
Factor Loadings for Principal Components Analysis With Orthogonal Rotation
Item Stable Internal
Acute External Acute Protective S4 Problem solving .73 .07 .10 -.24 S5 Entitlement .68 .09 .07 -.22 S3 Impulse control .67 .02 .19 -.21
S2 Attitudes towards authority .62 .18 -.05 -.31
S1 Peer associations .51 .18 .04 -.22 A3 Opportunity/access to victims .49 .31 .08 -.05 A4 Negative mood .05 .77 .12 -.07 A2 Anger/hostility .18 .68 -.25 -.02 A1 Substance abuse .20 .56 .12 -.10 A6 Interpersonal relationships .01 .06 .80 -.09 A5 Employment .40 -.07 .49 .04 A7 Living situation .06 .48 .49 -.18
S6 Attachment with others .24 .03 .47 -.38
P2 Prosocial identity -.24 -.03 .03 .75 P3 High expectations -.21 -.06 -.08 .75 P6 Social control -.17 -.05 -.13 .73 P5 Social support .01 -.22 -.34 .68 P4 Costs/benefits -.23 -.06 .04 .66 P1 Responsiveness to advice -.32 -.08 -.09 .59
Table 9
Correlations Between DRAOR Subscales Based on PCA Four-Component Solution
Subscale Stable Total Internal Acute Total External Acute Total Protective Total Stable Total 1
Internal Acute Total .36** 1
External Acute Total .43** .25** 1
Protective Total -.55** -.26** -.42** 1
**p<.01
Both the original (i.e., the structure as it was proposed by Serin, 2007) and new structures are presented in the figures on the following page. The most notable
difference between them is that the acute items split themselves up among three of the new subscales (Stable, Internal Acute, External Acute). One of the original stable items loaded onto the new External Acute subscale. The four-subscale DRAOR will be used for the remainder of this study (and thesis) because it is the best fit to this high-risk sample, although the results would still be quite similar if we used the original three subscales. The validity of the DRAOR total score (the sum of the risk factors corrected for the number of protective factors) will also be examined.
Stable Subscale Acute Subscale Protective Subscale Peer associations Substance abuse Responsive to advice Attitudes towards
authority Anger/hostility Prosocial identity
Impulse control Opportunity/access to
victims High expectations
Problem-solving Negative mood Costs/benefits
Sense of entitlement Employment Social supports
Attachment with others Interpersonal relationships Social control Living situation
Figure 15. Original Three-Subscale DRAOR Structure
Stable Subscale Internal Acute Subscale External Acute Subscale Protective Subscale Peer associations Substance abuse Interpersonal
relationships Responsive to advice Attitudes towards
authority Anger/hostility Living situation Prosocial identity Impulse control Negative mood Employment High expectations
Problem-solving Attachment with
others Costs/benefits
Sense of entitlement Social supports
Opportunity/access
to victims Social control