ELIMINATING THE EFFECT OF CONCOMITANT VARIABLES ON SALES DATA USING THE ANALYSIS OF COVARIANCE TECHNIQUE

Abstract

In marketing research, the understanding of relationships between variables by statistically controlling for extraneous sources of variation is refined by eliminating covariates. This work examines and eliminates the covariate effect on the sales of grand cereals oil. The data set for the work was on yearly sales (in millions of Naira) of grand cereals oil, Managers’ Age, Gender, and Operating Cost (the covariate) for a period of thirty (30) years (1991 - 2020). A general linear model (GLM) was first fitted to the sales data without the covariate. Then, another GLM with the covariate included (analysis of covariance (ANCOVA)) was fitted. Then, the proportion of total variability in the data accounted for by the covariate alone and by both the dummy variable and the covariate was investigated. The GLM result shows a highly significant effect of both the gender and age, with P< 0.0001, R2 of 49.8%, and mean squared error (MSE) of 70.782. The ANCOVA model showed a highly significant age effect with P< 0.0001, R2 of 80.5%, and a MSE of 27.028, indicating a sharp increase in the power of the test. The study has established the potential of the ANCOVA technique in controlling and eliminating the effect of the concomitant variable.