[robwwilliams]

Classic academic public relations piece. Not bad but more fluff than insight. Authors often have to grin and bear this PR machine, praying peers will forgive them their trespasses.

But here there’s a basic design flaw. This is a study of 16 ASD cases and 16 neurotypical controls. Small sample sizes like this require careful matching. The problem: the autistic subjects are 100% White but controls are 37.5% White. That imbalance can’t be waved away with statistics or Jedi mind tricks. Recruiting matched neurotypicals would have been straightforward.

One other issue is high heterogeneity within the two groups. In their Figure 1 (sorry behind a paywall), 4 - 6 of the autistic individuals have low mGlu5 levels across all regions. Two or three neurotypicals have high levels. Are these distributions actually normal, or are subgroups driving effects? It would help to know whether the participants’ GRM5 genotypes were informative wrt these subgroups. They weren’t checked.

[tgv]

Whenever you measure two different groups, you find a difference. Can it be solely ascribed to the one variable? Doubtful. You would have to match the groups for all possible contributing factors, and here not even the basic demographics have been matched? Another statistical effect for the irreproducibility bin.