The implicit bias test used is controversial, to say the least. According to mainstream cognitive psychology, it measures temporary perceptual associations via priming. These do not have a causal relation with higher level opinions. The effect can be caused by something as uninteresting as the local way of referring to science and scientists.
Methods? They had a large number of factors to correlate with their data: 25 (possibly a few more, depending on what you read), and ran a multiple regression over it, and are reporting an effect for every p
It's just another fishing expedition.