Confidence Interval Methods of Fixed Effects in Mixed Models: A Comparison Study

• Activity Number: 653
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Hatice Tul Kubra Akdur* and Deniz Ozonur and Hulya Bayrak
• Companies: Gazi University and Gazi University and Gazi University
• Keywords: Mixed models ; confidence interval ; fixed effect ; parametric bootstrap

Abstract

Inferences for fixed effects is usually main focus in mixed models. Many researchers suggest to construct confidence intervals of fixed effects or functions of model parameters to make inference. Parametric bootstrap methods to construct confidence interval are alternative methods to standard methods of inference and easy to apply when distributions of variances and pivots are unknown. We search for a promising confidence interval methods to make inference of fixed effects. Simulated data-sets of one-way random effect model fit to nested error regression model. For parameter estimation of fixed effect in nested error regression model, maximum likelihood based estimation methods is used. Assuming that maximum likelihood estimator is normally distributed and has known variance asymptotically, Wald-type confidence intervals are constructed for fixed effect. Bootstrap-t method is used to construct confidence interval of fixed effect. Furthermore, by modifying function of profile likelihood method, it is used to construct confidence interval of fixed effect. These confidence interval methods of fixed effect are compared in terms of coverage probability of true parameter value.