Читать книгу Elementary Regression Modeling - Roger A. Wojtkiewicz - Страница 16

Statistics courses in the social sciences most often begin with discussions of frequency distributions, cross-tabulations, central tendency, and variation. After some basics of probability and the normal distribution, students then are taught about the difference between means test (also known as the t test) and the chi-square test.

Regression analysis is usually introduced after this material is covered by discussing scattergrams and fitted lines. Although the preliminary material in the course involved comparing groups, the scattergram approach involves the relationship between two individual characteristics. The idea of comparing groups is absent. The disadvantage with using the scattergram approach is more than just statistical. A key part of social science thinking is thinking about how groups are different from one another.² The scattergram approach is not about groups but is about the linear relationship between different factors.

The discrete approach to regression modeling that I use in this book takes a different approach than the scattergram approach and builds on the group concept in explaining regression analysis. My approach first discusses dummy variable regression that connects directly to the difference between means test. The discrete approach thus builds directly on student knowledge about cross-tabulations and differences between means that is developed early in their statistics courses.