Lalonde, Trent

Committee Member

Yu, Han

Committee Member

Shafie, Khalil

Committee Member

Kole, James


College of Education and Behavioral Sciences; Applied Statistics and Research Methods


University of Northern Colorado

Type of Resources


Place of Publication

Greeley (Colo.)


University of Northern Colorado

Date Created



160 pages

Digital Origin

Born digital


The purpose of this dissertation was to establish measures that could be used to assess the relative fit of nested models with parameters estimated using the Generalized Method of Moments for longitudinal data with time-dependent covariates. A secondary data set collected from Filipino children was used as an example of model fitting to evaluate the quality of the assessment of fit of the Kullback-Leibler Information Criterion (KLIC) and a chi-squared statistic derived from the difference in the minimums of the quadratic forms of two candidate nested models. A simulation involving randomly-generated data sets was also used to evaluate the performance of the proposed statistics. Several variations of nested models were considered in the simulation, and the KLIC was used to compare the relative fit of these models. Overall, the performance of the KLIC as a model selection criterion showed that it achieved good detection proportion in identifying the correct model when it was compared to underfit models. On the contrary, it tended to favor overfit models over the correct model, and non-detection proportions were high when extraneous predictors were introduced to candidate models. Ignoring the feedback loop introduced by time-varying covariates and relying on the regular use of the Generalized Estimating Equations (GEE) for the analysis of longitudinal data could compromise model parameter consistency, efficiency, and bias resulting in misleading inferences. Replacing the former practice with the routine use of GMM to properly account for feedback in the data is highly encouraged. The KLIC would be a helpful tool to select an appropriate model among a collection of candidate GMM models, especially when there are time-varying predictors in the data.

Degree type


Degree Name


Local Identifiers


Rights Statement

Copyright is held by the author.