Linear models based on business data from the pharmaceutical industry

Abstract

This article describes the analysis of heterogeneous market data. For this purpose, the most relevant methodological aspects are discussed and analyses using a hierarchical linear model and multiple regression are presented. In the first step, the applied data set is presented, and the assumed hierarchical two-level structure is shown. The data are then prepared for the analysis. The data are checked for outliers, a multicollinearity check is conducted, a new variable introduced, missing values are replaced by estimated values, a transformation procedure is conducted in order to obtain normality, the data are aggregated for each hierarchical level and a sample size test is performed. The results of both methods are discussed. Finally, it is concluded that whereas the application of a hierarchical linear model appears to be one option, a multiple regression analysis can be employed instead if the quality of the data, especially the sample size, is not sufficient.