Computational protocol: Depressive Symptoms Affect Working Memory in Healthy Older Adult Hispanics

Similar protocols

Protocol publication

[…] The present study used CFA and SEM to model the hypothesized relations among affect, depression, and cognition and the covariates. The use of CFA and SEM allows for the modeling of true score variance by attenuating measurement error. Given the large number of indicators from the PANAS and GDS and the large number of parameters required to model them in CFA, parceling was used to conserve degrees of freedom and thus increase model parsimony. Parceling is an item aggregation method for creating parsimonious and just-identified CFA models [,]. Parceling has psychometric benefits in addition to parsimony such as improved reliability and closer approximations of multivariate normality []. Parceling is an appropriate method when the focus of study is on the structural model rather than the measurement model. We parceled the items by counter balancing the items based on their factor loadings in an initial model []. For the GDS, we used factor scores in order to form a more parsimonious construct and to have continuous indicators of the depression construct []. The factor scores were estimated from a model that deals with the binary nature of the items properly by using the WLS estimator with robust chi-square and standard errors [–]. The original items were grouped together into three factors scores, this factorial structure were based on an exploratory factor analysis that forced the indicators to 3 correlated factors. Full information maximum likelihood estimation (FIML, []) was implemented to handle missing data.The empirical validity of each model (i.e, how well the hypothesized model fits the observed data) was assessed using goodness-of-fit indices []. Model selection and evaluation was primarily based on differences in the root mean square error of approximation (RMSEA, []), the comparative fit index (CFI, []), the Tucker-Lewis index (TLI, []) and gamma-hat (γ̂) []. The RMSEA indicates the degree of mismatch between the sample variance-covariance matrix and the model-implied variance-covariance matrix with acceptable values being less than 0.10 []. The CFI and TLI are incremental fit indices that assess the extent to which the specified model improves fit over the null model, with acceptable values approaching one [,]. Lastly, γ̂ is a goodness of fit measure that has been found to yield unbiased estimates of fit in small samples, with values approaching one indicating better fit [].The SEM analysis first tested a full model, which simultaneously regressed all cognitive latent variables on all affect and depression variables, resulting in 12 estimated regression paths. In addition, this model regressed all latent variables on all covariates, resulting in 21 estimated regression paths. After testing the full model, the statistical significance of the contribution of each regression path was tested one at a time. This was done using iterative chi-square difference tests, where a model was estimated with one regression path constrained to zero and was then compared to the full model. Paths were retained when the chi-square difference, per one degree of freedom, exceeded the critical chi-square value, when α=0.01. Thus, not estimating parameters that fulfilled this criterion resulted in a significant loss in absolute fit and thus warranted their inclusion in the final model. Parameters not satisfying this criterion were excluded from the final model. This trimming methodology has been used by methodologists and is a robust alternative to trimming based upon the statistical significance of individual regression coefficients [,]. Analyses were performed with the package Latent Variable Analysis (lavaan) released 0.5.18 [] for the software R release 3.2.1 (R Core Team, 2015). […]

Pipeline specifications

Software tools MVN, lavaan
Application Miscellaneous
Organisms Cell fusing agent virus
Diseases Nervous System Diseases, Mitochondrial Diseases