Doctoral Dissertations

Date of Award

12-1994

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Human Ecology

Major Professor

Carl L. Dyer

Committee Members

Larry Wadsworth, Randy Bresee, Don Clark

Abstract

The purpose of the present study was two-fold: 1. To show disposable income as a sufficient and the more appropriate explanatory variable for total apparel expenditure. 2. To show that with appropriate redefinition/transformation of the data, the ordinary least squares method can be shown to be appropriate for apparel expenditure modeling. Three broad steps were used to redefine and transform the data. First, a heuristic experimentation was employed to obtain different functional forms for the three major quantitative variables, namely total apparel expenditure, total expenditure, and disposable income. The functional forms ranged fi*om the raw data forms, log-functional, and loglog-functional forms. Second, each functional form was divided by a functional form of the scaling variable family size. The optimal functional form resulting from the simple regression analysis was adopted as the form of the multiple regression model. The multiple regression model had the specified functional forms of either disposable income or total expenditure and of the number of income earners and family size. Three models, namely, the Income Model, Expenditure Model, and the Combined Model were formulated. The third and final analysis was the principal components regression intended to deal with multicollinearity in the Combined Model. This analysis was in two stages: 1. A principal components analysis of the nine raw (redefined) variables used as independent predictors in the Combined Model. This analysis resulted in nine principal components, which are uncorrelated linear combinations of the nine redefined variables. 2. A multiple regression analysis with redefined apparel expenditure as the response variable and the nine principal components, denoted as PRINl through PRIN9, as the explanatory variables. In the final analysis, this study has shown that when the functional form of apparel data is redefined to reflect apparel expenditure per family member, disposable income per family member, and number of income earners per family member, and if the variables are transformed by loglogging all quantitative variables, subject to constraints on the scaling variable family size, disposable income is a very powerful predictor, and indeed accounting for multicollinearity, a slightly better predictor than total household expenditure, of total apparel expenditure.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS