Investigating the Behaviors of M₂ and RMSEA₂ in Fitting a Unidimensional Model to Multidimensional Data.

ABSTRACT: It has been widely known that the Type I error rates of goodness-of-fit tests using full information test statistics, such as Pearson's test statistic χ² and the likelihood ratio test statistic G², are problematic when data are sparse. Under such conditions, the limited information goodness-of-fit test statistic M₂ is recommended in model fit assessment for models with binary response data. A simulation study was conducted to investigate the power and Type I error rate of M₂ in fitting unidimensional models to many different types of multidimensional data. As an additional interest, the behavior of RMSEA₂ was also examined, which is the root mean square error approximation (RMSEA) based on M₂. Fin

SUBMITTER: Xu J 

PROVIDER: S-EPMC5978478 | biostudies-literature | 2017 Nov

REPOSITORIES: biostudies-literature