Analysis of covariance using Python

This is the third video lecture in my seminar series on linear models. Here, I discuss analysis of covariance (ANCOVA). We combine what we have learned about linear regression and analysis of variance.

In ANCOVA we have a categorical variable as independent variable and a continuous numerical variable as dependent variable. This is typical in a study design, where participants are assigned to the levels made up by the unique values of the independent variable. Other variable might also have an effect on the dependent variable. We might not be able to control for these with the study design. If such a variable is a continuous numerical variable, we term it a covariate. So, in ANCOVA we look at the effect of the categorical independent variable on the continuous numerical variable after correcting for the covariate.

There are important assumptions for the use of ANCOVA models. The index below shows all the sections of the video lecture.

Please watch the first two videos in this seminar series, as I use concepts explained in these two videos in the tutorial on ANCOVA.

Linear regression: https://www.youtube.com/watch?v=6gtEfIbx6ko
ANOVA: https://www.youtube.com/watch?v=yeRtgKtbUFU


00:00 Introduction
05:55 Import data
20:23 Linearity
22:02 Homogeneity of regression
24:56 Normality of residuals
26:43 Homogeneity of variance of the residuals
27:53 Outliers
29:16 ANCOVA