Methodology for estimating missing information and ANOVA correction with the cell means model in connected designs

Diana C. Franco SOTO[1]

Oscar O. Melo MARTÍNEZ1

§     ABSTRACT: In this paper we present a non-iterative methodology to estimate missing data in connected cell means designs. This methodology improves on that of Melo (2002), reducing the correlation between observed and estimated information. After imputing the missing information, we suggest a way to analyze the variance by accounting for the subvector's covariance structure of the information vector. This analysis guarantees that the test statistic follows a central F distribution under H_0. Finally, we work on a case study of a factorial design, in which the proposed methodology is applied.

§     KEYWORDS: Cell means model; connected designs; missing information; estimation and imputation; distribution of quadratic forms.



[1]Department of Statistics, Universidad Nacional de Colombia, Bogotá, D. C., Colombia. E-mail: dcfrancos@unal.edu.co / oomelom@unal.edu.co