Apr 20, 2018
[Content warning: harassment. This discusses the comments to SSC Survey Results: Sexual Harassment Levels By Field]
Thank you for posting this and the data file. FWIW, I tried to reproduce the results and couldn’t reproduce the correlations between female victimization, male victimization and male perpetration. fem vic vs. male vic is 0.65, same as yours. fem vic vs. male perp is 0.01 for me, and male vic vs. male perp is 0.21 for me. Everything else more or less checks out.
As a reviewer, I’d say the combination score is not convincing, especially since it ignores all considerations of different male to female ratios in the various industries.
Also, if you have two measures with r = 0.8, Fig 6 is not a good idea IMHO. It’s probably just noise. (Also, it should be a dotplot centered around 1, because the relevant info is distance from 1:1 ratio.)
Instead, I’d focus on the correlation between female victimization at work and female victimization outside work of 0.65 (for me) and the same for males at 0.59, which also leads to the conclusion that there’s a strong ‘people in fields’ effect, without having to go through the combination score. If you’re so inclined, you might then do the at-work by outside-work ratios and end up a kind of cross-validation set, where you can see whether the bad fields for women are bad for men as well. Of course, once you then consider sex ratios per field. it’s story time all over again. Still, e.g. men report similar levels of out of work victimization in computers (20%) and Health Care (24%), but at work victimization of 4% and 12% respectively, which strongly suggests that Health Care is worse.
Their code is available here. Thanks for doing the work to try to replicate my results. I’ve removed the non-confirmed correlations from my post until I can figure out what’s going on with them. I agree that Figure 6 was barely worth it, which is why I tried to make Figure 4 (the unadjusted version) the center of my thesis.