This component contains the materials required to conduct our supplemental analyses, as well as some results from those supplemental analyses.
1. **institute_and_race.R** contains the R code necessary to test whether the degree of pro-White-male bias varies by broad scientific topic area (embodied by the institute that funded each grant) and by whether the reviewer was a White male
2. **check_variances.R** contains R code fit multi-group, multi-level Structural Equation Models to test whether the random by-reviewer and by-proposal intercepts differ by PI race and gender. Thanks to Josh Pritkin for his help implementing these models (https://openmx.ssri.psu.edu/node/4399)
2. **word_counts.R** contains the code required to conduct our word count analyses. Fitting these models takes some time, so fitting these models follows a similar workflow to our sensitivity analyses: using resources from the Center for High Throughput Computing (http://chtc.cs.wisc.edu/), I passed names of each word count outcome variables to each of nine computers. Each computer uncompressed a file containing an R installation (R.tar.gz) and a small shell script (runR.sh), which instructs the computer to run word_counts.R. Results are then passed back to the central computer
3. **word_count_graphs.R** contains code to make our figure of the word count results
4. **Word count results table.csv** contains model fixed effects for all the word count analyses