DUMMY VARIABLES AND EXPERIMENTAL DESIGN X11 Dummy variables were introduced in Section 12.8 in applications involving regression models applied to two discrete categories of data. For example, we saw how they could be used to test for gender discrimination in the salary example, In this section we expand the potential applications of dummy variables. First, we present an application in which a regression model is applied to more than two discrete categories of data, Next, we show how dummy variables can be used to estimate the sea- sonal effects on a regression model applied to time-series data. Finally, we show how dummy variables can be used to analyze data from experimental situations, which are defined by multiple-level categorical variables. We also provide an example that shows how dummy variables can be used for public policy analysis. Example 13.1 Demand for Wool Products (Dummy Variable Model Analysis) A senior marketing analyst for the American Wool Producers Association is interested in estimating the demand for wool products in various cities as a function of total dis- posable income in the city. Data were gathered from 30 randomly selected Standard Metropolitan Statistical Areas (SMSAs). As a first step the analyst specifies a regression model for the relationship between sales and disposable income: Y = Bo + Box where X, is the per capita annual disposable income for a city and Y is the per capita sales of wool products in the city. After some additional discussions, the analyst won- ders if overall sales levels differ among different geographic regions: north, central, and south