Replication Materials for Lawcourt-L post


Here is the SPSS output which includes commands used to generate the logistic regressions.  The first set of results is for dec_types 1, 6 and 7 and the second is for all dec_types.  As you can see, the PRE is lower in the first instance.  Including memorandum cases, it seems, drives the PRE up, mostly because the null predicts better in cases without memorandum cases included (e.g., there is less variance available to reduce, so the model reduces less variance) and the model predicts a bit worse as well. 


As a side note, 23% of votes in the individual level databases (including Warren, Burger, and Rehnquist Courts) are memo cases and (many more than I assumed), and, of those, 75% are conservative in direction.


Click here for the SPSS output.


Click here for the merged dataset (or download it from The Ulmer Project at the University of Kentucky).