And just from visual inspection, it is evident that there are numerous surprises with the iSeeCars study that are difficult to explain. For example, why is there such a huge difference between Model 3 and Y fatality rates even though the cars are similar in most respects?

Another thing to note is that the number of fatalities is only a handful, often under 10 in a year. The standard deviation of the Poisson distribution would already give error bars of 30% at least, making direct comparisons between models meaningless.

Finally, if you read the accident reports, most of the fatalities are head on collisions with drunk drivers and not wearing seatbelts. Having physical controls for hvac or whatever has absolutely nothing to do with it.