Listwise versus pairwise missing values
As was shown in the first correlation matrix earlier in the chapter, missing values are, by default, handled in a pairwise manner in the correlation procedure. In linear regression, the default is to exclude cases on a listwise basis. While the default setting can be changed in each of these procedures, it is important to appreciate the differences and decide how to best handle this aspect of the analysis process.
Note
The pairwise approach makes use of all the available information, so it provides the most complete picture of the linear relationship between a pair of variables. Listwise handling of missing values creates a matrix in which each coefficient is based on the same set of observations, which provides consistency.
Comparing pairwise and listwise correlation matrices
When a set of variables is going to be used in a regression analysis, it is a good idea to use correlations to assess all the bivariate patterns, and part of this evaluation involves...
