A Statistical Tool: Analysis of Covariance. Volume III. Program Listing, Flow Chart, and User's Manual for Algorithm for Handling Multivariate Covariance Data with Missing Values.

Abstract

Abstract : This report reviews the theoretical framework which permits analysis of multivariate covariance data in which missing values occur among both dependent and independent variables. A flow chart and program listing is given for a computer program which will estimate the block and treatment parameters, as well as the regression coefficients for the covariates. The program will also calculate test statistics for testing hypotheses which are supplied by the user. It should be pointed out that this program allows for only two categories of design indicator variables, namely a block effect and a treatment effect which are assumed to be additive. The program is thus suitable for the usual randomized block model with additive block and treatment effects, or a general two-factor additive effects model (i.e., no interaction), or the usual one-way classification model (i.e., several treatments but only one block). It will not accommodate more complex designs such as a Latin Square Layout with a multivariate response and one or more covariates. User instructions and a worked example are provided in this volume. (Author)