6.4 Strengths and Weakness of MLR Models
Multiple linear regression as a modeling method has the following strengths and weaknesses.
Strengths of the MLR Method
-
Models are transparent (interpretable). How input variables contribute to the value of outcome variables is explicity defined.
-
MLR helps you identify useful predictors are part of the modeling process itself. Tools that you used to develop MLR models often will provide a step-wise option that shows you which variables contribute most to prediction. Also, the tools show explicitly whether the overall model is statistically significant and whether each coefficient is statistically significant. When coefficients are not statistically significant, you can experiment with dropping the predictors to see if your power to predict stays the same, gets better, or gets worse.
-
Works well when relationships between predictors and outcome variable are approximately linear.
Weaknesses of the MLR Method
-
Can proliferate variables through dummy variables.
-
Does not work well when the relationships between predictors and outcome variable are not approximately linear.