ABSTRACT
B. The PPM with correlated errors between subdomains had an acceptable fit (χ2 (df) = 115.650(39), CFI/TLI = .981/.968, SRMR = .052, RMSEA = .066). All between-subdomain correlated errors were statistically significant. Compared to the baseline PPM (without autocorrelated errors), the chi-square statistic was improved (Δχ2 (df) = 843.477(12), p < .001). However, the PPM with correlated errors within subdomains had a poor fit (χ2 (df) = 940.562(42), CFI/TLI = .777/.650, SRMR = .069, RMSEA = .218), and most within-subdomain autocorrelated errors were not statistically significant. Although the chi-square for this model was also significantly improved compared to the baseline PPM (Δχ2 (df) = 18.565 (9), p < .05), we excluded this model from subsequent analyses due to its overall poor fit. For these reasons, the PPM with correlated errors between subdomains was selected as the optimal model.
