Seminar

Karabi Sinha

The Effect of Measurement Error in Analysis of Covariance [ANCOVA] Models

Analysis of Covariance [ANCOVA] models are known to possess the special characteristics of blending the twin features of linear models for traditional [varietal] designs on one hand and the regression designs on the other. The design matrices are typically matrices involving the incidence patterns of assignable causes such as block effects or row-column effects, apart from the varietal or treatment effects. The incidence of such effects is reflected by the binary nature of the incidence matrices. On the other hand, the regression matrices reflect the extent of the regressors which are generally non-stochastic and continuous in nature.

In this work, we propose to develop the general theory and related data analysis techniques for a covariates' model in situations wherein some or all of the covariates are subject to measurement errors.

In particular, we confine to a Completely Randomized Design set-up and present detailed results from power considerations involving one covariate with measurement errors in a structural model.

Seminar Date:
January 31, 2007