R Part 5

The total length of the videos in this section is approximately 43 minutes, but you will also spend time running code while completing this section.

You can also view all the videos in this section at the YouTube playlist linked here.

Please download the two code files below in order to follow along with the videos.

Graphically checking whether logs would help

Question 1: Which assumptions can be made more true by logging one or both variables?

Voltage example

Question 2: What happens if you have a bunch of dots along a curve, and you connect them in the wrong order?

Bands, and how to make and interpret residual v. fitted value plots

Question 3: When are residual v. fitted value plots most useful?

predict function for linear regression

Question 4: Given a linear regression, if you don't specify a new data set to make predictions for, or if there is an error in your specification of a new data set, what does the predict function do?

predict function for trees and splines

Question 5: If you have a spline instead of a tree or lm, what is the "new" argument called?

That's it.

During this tutorial you learned:

• What a scatterplot of data will look like if the logging the outcome variable (Y) and/or predictor variable (X) would lead to a data set that better meets meet the assumptions of a linear regression, including linearity, equal variance, and/or normality

• How to plot the fitted values for each value of x with points() and abline()

• Two ways to plot the confidence intervals and prediction intervals for a linear regression model via calculation or the predict() function

• How to plot residuals versus fitted values and how to use the residuals v fitted values plot to check assumptions of a linear regression model

• How to use the predict() function and tips to troubleshoot common problems

Terms and concepts:

Log transformation, confidence interval, prediction interval, residuals versus fitted values plot

Functions in review:

exp(), log(), qt(), order(), predict()