| Now Structural Equation modeling (SEM) has become a versatile and widely used data analytic method for evaluating causal and predictive hypothesis in Medical research. By far the most common method used to estimate SEM is normal theory ML. Two key limitation of ML estimation is the strong assumption of multivariate normality and large sample size. However ,applied researchers often find themselves with data clearly violating the distributional assumption ,and without sufficient sample size to utilize some distribution free estimation methods.So, there has been increased demand for methods that perform optimally at smaller sample sizes and under varied distributional conditions.Two another methods were introduced to solve problems above:Satorra-Bentler(S-B) scaled and bootstrap resampling methods. They are respectively offered in EQS and AMOS software package. Monte Carlo computer simulations were used to investigated the performance of three estimation methods:ML,S-B and bootstrapping .They were examined under varying conditions of sample sizes (N= 100,250,500) and multivariate distributions (normal,moderate and severely departure from normality) ,as well as using a practical example. Using a more valid estimating methods would show that our conclusion using existing methods would be clearly in error.For properly specified models,ML showed no evidence of bias under normal distributions across all sample sizes. But ML test statistic was increasing overestimated with increasing nonnormality,and its standard errors to parameters were underestimated. Severe distortions in all aspects of the final solution can result. S-B scaled ( but at the smallest sample size) and bootsrapping methods were relatively stable with both normal and nonnormal data.Still .through continued research such as the present investigation ,SEM techiniques continue to evolve ,breaking sample size and distributional barriers,thus becoming more accessible and useful to practitioners.
|
PDF Full Text Request