General Linear Model Specification Error Test with Missing Data

In this paper, we consider a general linear model where missing data may occur in response and covariate variables. We propose a new test based on Ramsy's test to identify goodness of fit for general linear model with missing data. We show that under the null hypothesis, our test functions for complete case analysis follow a Fisher distribution and the other test function used for analysis with available data converges in distribution to Quasi-Fisher distribution. Furthermore, we compare proposed test functions by using some simulation studies. Also, we apply our methods in analyzing a real data set.