TIME-VARYING COEFFICIENT MODEL COMPONENT ESTIMATION THROUGH GENERALIZED ESTIMATION EQUATIONS

Abstract

A methodology to estimate time-varying coefficient model's components through generalized estimation equations (Liang & Zeger 1986) is proposed, in order to include directly in the estimation the possible correlation between repeated measurements of each subject. Expansion of the time-varying coefficients is done by means of regression spline methods (Huang et al. 2002). Furthermore, is proposed the use of the Akaike's information criterion in generalized estimating equations (QIC) proposed by Pan (2001) like model selector. Through simulation are compared the proposed methodology and the methodology presented by Wu & Zhang (2006), where model's components are estimated through weighted least squares and Akaike's information criterion (AIC) is used like model selector. It resulted that the proposed methodology gives a better behavior in relation with the average mean square error. In order to illustrate the methodology, is taken into account the data base ACTG 315 (Liang et al. 2003) related to a AIDS study, where it is investigated the relationship between the viral charge and the CD4+ cell count.