This recipe will look at studying a multiple regression on determining the sleep duration of mammals. The dataset is available at StatSci.org.

We will run the study with all predictors included for the initial model and then remove the terms in the model step-by-step. Our goal is to reduce the terms until only the significant ones are left. As this data does have a degree of correlation among the predictors, we will use matrix plots, correlation, and variance inflation factors to highlight the degree of multicollinearity.

We will initially produce the matrix plots and the correlation scores before moving on to the analysis of the regression.

After reducing the model to only significant terms, we will then produce the residual plots to gradually verify assumptions on the analysis.

The value of alpha for the decision level used here will be 0.05.