Title: A semiparametric mixed analysis of covariance model for a crossover design with carryover effects
Authors: Leonard Allan Almero - University of the Philippines Los Banos (Philippines) [presenting]
Erniel Barrios - University of the Philippines (Philippines)
Joseph Ryan Lansangan - University of the Philippines (Philippines)
Abstract: A semiparametric mixed analysis of covariance model for a crossover design with carryover effects is postulated. The responses are adjusted for covariate effect through a nonparametric function of the covariates. To estimate the model, a hybrid of restricted maximum likelihood estimation and smoothing splines regression is imbedded into a backfitting algorithm. A bootstrap-based test is then developed for testing differences in treatment means for fixed effects and/or significance of variance components for random effects. Simulation studies indicate that for a random effects model, the bootstrap-based test for variance components is correctly-sized. The test is also powerful and relatively robust to the hypothesized magnitude of variance. For a fixed effects model, the bootstrap-based test performs relatively better than the ordinary analysis of covariance when under 10\% mean effect differences. Furthermore, for a model with either a fixed effects or random effects, the test remains advantageous over ANCOVA in the presence of misclassification error and in non-normal error on unbalanced data.