Use Of Latent Variable Measurement Model And Classical Test Theory Analysis Of Reliability And Validity Study

The author has done three studies on the basis of related researches using practical data and simulative data to identify the differences in reliability and validity studies between Latent Variable Measurement Model ( LVMM) and Classical Testing Theory ( CTT ). The first one is to compare the reliability and validity analysis when applying LVMM and CTT; the second one is to find out how correlated errors in LVMM can affect parameter estimation; and the last one is to compare the different results of construct validity when LVMM and CTT are applied to the Multitrait-Multimethod design (MTMM).The studies show that, weak hypotheses are assumed in LVMM; the reliability and the validity can be estimated directly and systematically with LVMM; and the results are more reliable and reasonable. The studies also show that, in the process of fitting LVMM, correlated errors really can affect parameter estimation regularly and the effect can be successfully reduced; when the correlation among errors is lower, error variance has less influence to the fitting parameter, but the influence will increase as error variances become larger; and when the correlation of errors and error variance are much higher simultaneously, the correlated errors have the biggest effect on fitting parameters. Using LVMM to analyze construct validity in MTMM can clear up the logistic errors caused by CTT.These studies provide a useful technological support for psychological and educational researchers to get more reliable summaries. And the studies also provide a method to avoid the influences caused by correlated errors in the process of Model fitting.