The Panel Data Case
If we remember from Chapter 3, variance heterogeneity can rarely be detected with a test for homoskedasticity. To reiterate equation (3.5) from that chapter, we have
r 2 = c0 + c1D + m, (8.1)
which tells us that in the case of variance heterogeneity, a correlation would exist between the squared residuals and a set of dummy variables representing a particular clustering of the data, much like in intercept and slope heterogeneity. However, we run into an issue if we perform a regression specified exactly like (8.1). That is because the dependent variable is a squared version of a normally distributed variable (at least we hope it’s normally distributed). Park (1966) recommended using the natural ...
Get Building Better Econometric Models Using Cross Section and Panel Data now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
