9.2 Assumptions of linear models

From last week:

Now that you hold real power in your hands to do data analysis, we need to to have our first talk about due diligence and assumptions of the statistical models that we use. There are three fundamental assumptions that we either need to validate or address through experimental design in this class of models.

1. Independence of observations.
2. Normality of residuals (with mean=0)
3. Homogeneity of variances (i.e. homoscedasticity)

We will discuss what each of these means in class this week, and during the next several weeks we will discuss methods for verifying these assumptions or relaxing the assumptions to meet our needs through specific techniques.