145Extensions
a mean of zero and variance of ν
k
(Hedeker and Gibbons 1994; Raudenbush and Bryk
2002; Rabe-Hesketh and Skrondal 2012).
The second column in Table 6.4 presents results from a multilevel random slope
model using the same data from the previous example. In this model, both the inter-
cept and the slope for logged family income are allowed to vary across countries, and
the Level 2 errors are unstructured (i.e., the random coefcients are allowed to be cor-
related). Results for the regression parameter estimates at Level 1 are very similar to
those from the random intercept model, except that the magnitude and signicance
level of parameter estimates appear to be smaller for each variable. This is not surpris-
ing because as we in