Gender bias in student evaluation of teaching or mirage?

In a recent small sample study, Khazan et al. [1] examined SET ratings received by one female teaching (TA) assistant who assisted with teaching two sections of the same online course, one section under her true gender and one section under false/opposite gender. Khazan et al. concluded that their study demonstrated gender bias against female TA even though they found no statistical difference in SET ratings between male vs. female TA (p = 0.73). To claim gender bias, Khazan et al. ignored their overall findings and focused on distribution of six “negative” SET ratings and claimed, without reporting any statistical test results, that (a) female students gave more positive ratings to male TA than female TA, (b) female TA received five times as many negative ratings than the male TA, and (c) female students gave “most low” scores to female TA. We conducted the missing statistical tests and found no evidence supporting Khazan et al.’s claims. We also requested Khazan et al.’s data to formally examine them for outliers and to re-analyze the data with and without the outliers. Khazan et al. refused. We read off the data from their Figure 1 and filled in several values using the brute force, exhaustive search constrained by the summary statistics reported by Khazan et al. Our re-analysis revealed six outliers and no evidence of gender bias. In fact, when the six outliers were removed, the female TA was rated higher than male TA but non-significantly so.