Use of Ordinal Dummy Variables in Regression Models

Abstract

Many activities and phenomena on earth which are of interest to man and require to be studied are not quantitative in nature, they are rather qualitative.Sometimes their contributions, as independent variables, in a multiple regression, to variations in a specified quantitative dependent variable and to characterize them are of interest.Usually dummy variables with equal spacing are used in generating the design matrix despite its uninterpretable coefficients.Here a cumulatively coded design matrix is proposed and the coefficients are interpreted.This method is applied to an illustrative example, alongside two other possible methods, to demonstrate its applicability, and the proposed method showed a comparatively good performance with an additional advantage of interpretable coefficients which is very useful for practical purposes.