Return to Previous Page
In regression analysis, the use of residuals can lessen or neutralize the effects of confounders. Residuals are essentially what is remaining when a model has been run to control for several potential confounding variables. The possible confounders are first entered into a regression model and the residual is what is left over to account for everything else. The idea is to save the residual as a new variable after running the model with the suspected confounding variables (this is an option in SPSS under REGRESSION). The next step is to run the model again with the variable of interest and other variables that are not suspected confounders using the RESIDUAL as the new DEPENDENT VARIABLE.
This is most useful in multiple regression analysis and will be used in the following section as an example for controlling for the confounding of age in a model with measles immunization and other factors in predicting nutrition status. Here age (suspected confounder) and other possible confounders will be entered into a linear regression model and the residual will be saved as a new variable. The residual contains the attributes that are not influenced by age and the other entered variables. A new model is run entering measles immunization and other variables (not confounders) using the RESIDUAL as the dependent (outcome) variable to control for the effects of the confounders.