The Gender Paradox in Comorbidity between Depression and Aggression: A Longitudinal Invariance Analysis in Young Adolescents
Description
Metadatos from AMOS program
Files
Steps to reproduce
This study is based on the conception of a model composed of three first-order latent factors: depression, physical and verbal aggression, which share a common underlying risk factor or second-order latent factor. Analysis of the longitudinal invariance of the proposed model will allow us to know whether there is equivalence between adolescent boys and girls in how they are affected by the underlying factors common to the comorbidity between depression and physical and verbal aggression. In this regard, we will follow Byrne's recommendations (2016), who indicates a procedure based on the evaluation of invariance at three levels (configural, metric, and structural) based on the different restrictions imposed on the parameters that define the base model. Specifically, the so-called configural invariance model attempts to demonstrate whether the factorial structure of the model is plausible and, therefore, is shown to be an adequate base model without any restrictions. If the configural invariance fits the data, the second step tries to prove that there are no gender differences due to the psychometric instruments used through the so-called metric invariance. At this second level of invariance, the equivalence between groups is analysed by restricting the factor loadings or weights of the model. The next and last step correspond to the evaluation of the structural invariance at the level of latent variables. The invariance of the variances of the factors between groups is analysed. At this level, the specification of the model will include, in addition to the equality in the factor loadings, also those of the variances in physical aggression, verbal aggression and depression. To conduct a comparative analysis of each of the two invariance models (metric and structural), which we could refer to as nested since both derive from the base model (configural invariance), the increase in degrees of freedom of the chi-square index or maximum likelihood ratio test will be used. If this value is statistically significant in either of the two nested models, it means that the specified restrictions in the more restrictive model do not hold and, therefore, the model is not equivalent between boys and girls at that level of analysis (Barrera et al., 2015). Additionally, due to the sensitivity of the chi-square to sample size and the assumption of normality, the increase in Comparative Fit Index (CFI; Bentler & Bonett, 1980) is also indicated to determine the equivalence of the nested models. When this increase is greater than 0.01, the least constrained model is accepted, and the other is rejected. Likewise, the fit indices Root Mean-Square Error of Approximation (RMSE; Steiger, 1990) are also presented, as indicated as appropriate for evaluating the validity of this type of analysis (Mai et al., 2021). RMSEA requires a value less than.05 for the acceptance of the proposed model, representing reasonable errors of approximation of the model to the population.