Two items (28 and 30) misfit the model's expectations based on fit residual values above 2.5, and five more items (7, 8, 9, 21 and 26) also had a significant chi-square statistics or F-statistics (Table 2).
Based on the fit statistics obtained, it was determined that the three-class model provided a better fit for these data.
Based on these fit statistics for the original model, we elected to re-specify our original model to improve its fit.
The final model was selected based on model fit statistics, including the Akaike Information Criterion AICC), the Bayesian Information Criterion (BIC), the Root Mean Square Error of Approximation (RMSA), the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI) and the Standardised Root Mean Square Residual (SRMR), as suggested by Hu and Bentler (1999).
Misfitting items are deleted starting with the most misfitting, based on their fit statistics or the presence of DIF.
Last, multivariate models were selected based on model fit statistics (Akaike information criterion and r) and clinical significance.
A lag of 0-1 days was chosen for principal analyses based on model fit statistics from preliminary analyses, but lags of 0 through 7 days were tested.
So the confidence intervals, based on goodness of fit statistics, were very narrow.
Number of classes determined based on model stability, fit statistics, and interpretability of latent classes Abbreviation: CAM, complementary and alternative medicine Participants were classified into latent classes corresponding to their highest posterior probability, as derived from the LCA.
For the analysis using daily PM2.5, a 1-day lag provided a slightly better fit, based on t-statistics, than either lag 0 or lag 2. With the less frequent PM2.5 data collected every 3 6 days, lag 2 provided the best fit.
