The Method and Use of Novacom, a Program for 'Non-Orthogonal' Analysis of Variance and Covariance

Abstract

The report contains the description of a program ("NOVACCI") for the solution of problems in the area of analyzing "data based on the general linear statistical model. While the detailed program documentation is riven elsewhere, the present publication deals with the statistical method, the logical flow, and the use and application of NOJACC4 in multiple linear regression and (" non-orthogonal") analysis of variance and. covariance for crossed classifications with incomplete and unbalanced data. The method of NOVACC14 is basically a backward ranking procedure applied to individual and/or groups of independent variables (concomitant independent variables and/or AINOVA effects, respectively). The result o-L the ranking is a model ("significant model") which contains only significant concomitant independent variables and/or AflOVA effects. The method and use of the program is illustrated by examples of the statistical analysis of bodies of incomplete experimental data.