A comment from SBM
"In the Abstract, Methods, and Discussion, Hooker says he used Pearson’s chi squared test. But all four data tables are labeled Fisher’s exact analysis. I’m pretty sure those are not the same test (e.g. see this SAS page). That’s a pretty glaring error! It suggests that Hooker either didn’t realize that the tests are different, or consistently mislabeled his Tables and never noticed, or perhaps that he evaluated the data with both tests and was too careless to ensure that the final manuscript was consistent.
Hooker also claims, in the Methods, that the Pearson’s chi squared [sic] results were confirmed by “a conditional logistic regression design similar to the Destefano et al. If so, why didn’t he bother to actually show that result? It’s not as if he was under a strict page limit.
Next, I note that Hooker didn’t indicate the number of subjects being analyzed in each group for each of his tables. In fact, the original DeStefano publication says that 93.4% of cases were vaccinated by 36 months. That means that there were only 41 total children with autism that were in the >36 month group. If 20% of those were African American, that’s only 8 kids.
Most important, of course, is that even if this association is real, is there any good reason to think it’s causal? Hooker certainly doesn’t provide one. In particular, why is the association strongest in children who received their first MMR after 36 months, weaker in the >24 group, and non-significant in the >18 group? That makes little sense if MMR is actually a causative factor. It would mean that MMR only causes autism in African American boys, and even then, only if the vaccine is NOT given by 18 months!
I look forward to Hooker calling for a concerted campaign to ensure that all African American boys receive MMR before 18 months, to prevent them from developing “late-MMR-associated autism