I just think the arithmetic mean is weird to use, because of the way we think about selection coefficients. They are relative increases in fitness, and so work on a multiplicative way. A mutation of s=0.1 increases fitness by 10 percent, and a second mutations with s=0.2 by another 20 percent (total w=1.1*1.2), so it just feels strange to use the arithmetic mean for the total fitness.

But whatever I feel about it, it’s very hard to measure individual fitness components, which you would have to do to test it.

]]>I put in the arithmetic mean mostly because I didn’t really think it through and have no opinion one way or another. For average effects the arithmetic mean is what Fisher (and Falconer) use. So, I am open to being corrected on that, but you would have to tell me why something else would be a better choice.

By the way, I can’t figure out the keys I need to spell your name correctly.

]]>