Epidemiology question. When epidemiologists estimate "transmissibility" of a viral variant, are they estimating the observed transmissibility or estimating something causal about the mutation? @CT_Bergstrom @noamross

Imagine a virus that undergoes random mutations at some rate and that none of these mutations change anything about the virus or how it spreads. Some of these mutations would arbitrarily spread to fewer people and some to more.

I suspect that the distribution of variants would be non-linear with most variants infecting few people and a few infecting many people. The variants infecting more people would also be more likely to be discovered and studied.

If you estimated the observed transmissibility of the variants that happened to spread the most, they would be higher than the baseline rate - but this is just due to random chance + sampling bias. There is actually nothing causal about the mutation.

Do epidemiological techniques compare observed transmissibility to what you would expect from a null model like the one above? Or do they just try to calculate the average number of people infected without a correction for sampling bias?