Learning Log

Today in class, we covered more materical from Chapter 5. We talked about diagnostic plots, transformations on variables and outliers.

We talked about plotting in R and the shortcuts that come with it. When you use the plot command in R, you see four plots pop up.

The first one is the residuals vs. fitted values. This one we would ideally like to see no trend. We want all of the resiudals to be equalls spaced apart. We see an “ideal” line and an “actual” line. We want these two lines to match up.The most important thing is that We want there to be NO curvature. If there is curvature, this means the mean function is missing something! We can also look out for heteroscedasticity.

In the second plot, which is the qqnorm plot, we ideally want all the points to be on the line.

In the 3rd plot, we are plotting the scale-location plot. We want to look for a trend in the residuals and skewness of the data. Ideally there would be no trend.

The las plot is the Cook’s distance plot. We want the points to be inside of the red lines. This is the ideal plot. This plot measures the influence each data point has on the regression coefficient.