Читать книгу Elementary Regression Modeling - Roger A. Wojtkiewicz - Страница 11

Control modeling involves first estimating a regression equation with particular emphasis on a coefficient for one of the independent variables in the analysis. I refer to this independent variable as the “independent variable of interest.” The next step is to estimate one or more regression equations with additional independent variables called “control variables.”

Adding the control variables ensures that the effect of the independent variable of interest does not capture the correlated influences of the control variables. Additional models may be estimated in an attempt to “explain” the initial coefficient of the independent variable of interest.

The need for control modeling arises when we analyze secondary data such as data from nationally representative surveys. When analyzing group differences, membership in one group may be related to membership in another group. Control modeling narrows the meaning of the coefficient for the independent variable of interest by removing related influences.

For example, if the regression analysis shows that children in stepparent and single-parent families score lower in academic achievement tests, the next step in the analysis is to control for family income. Family income would be controlled since it is related to both family structure and achievement.