", per the observation that there is evidence of discrimination against women when gender is identified."
Not sure how they come to this conclusion when they indicate that when the gender is identified, BOTH genders see a significant drop and men see a *greater* drop when they're known to the project. It's only when the women are unknown that their acceptance rate is lower... but even then, the acceptance rate of men and the acceptance rate of women's error bars overlap... it's entirely possible there's no difference between the genders when the contributor is unknown.
In fact, the only place in their pull request acceptance rate error bars don't overlap on p15 is where identified male insiders are rejected at a greater rate than women.
"We hypothesized that pull requests made by women are less likely to be accepted than those made by men."
Seems like bad research... start with a hypothesis and highlight areas of your study which weakly support it, ignore areas which strongly refute it.