Readings: Snijders and Bosker (2012, Ch. 2 & 3.1.-3.4).

Lecture slides

Web exercise 3a

Clustered data violate an important assumption of OLS regression: To the degree that observations in a cluster share similarities, their residuals are correlated. As a result OLS statistical inference can be overoptimistic, so that we falsely report results that are simple chance-findings.