2

I am studying a population of individuals over many generations. Each individual can take on a single trait from a range of several traits. In each generation I calculate the the share of the most common trait in the population. The result looks something like this:

enter image description here

Now I would like to say something along the lines of "the population has converged to a single trait at time T". To do this I have been using this naive notion of "convergence" where I simply take the first time point where the share of the most common trait at the time exceeded 95% (With no claim that this "convergence" has anything to do with convergence in the mathematical or statistical sense). I do this because I know from evolutionary biology (under a certain set of assumptions), that once this happens the odds of a new invasive trait being introduced via mutation that could take over the population are essentially zero.

I have been told by my colleague that I could improve my claims by pointing to convergence in distribution (or possibly other types of stochastic convergence). However, I don't see how I could show that my variable converges to any specific distribution (and what that distribution would be)? For starters my "trials" are not independent: the share in time T depends on the share in time T-1. Is it valid to even consider distribution in convergence in this case?

Peter R
  • 31
  • 1
  • Convergence in distribution can certainly be applied to a sequence $(X_n)$ of dependent variates. The question is rather whether or not you can establish this convergence. A graph does not tell anything about this. – Xi'an Oct 19 '16 at 19:26
  • I did not know you can apply it to dependent variates, thanks for that comment! And yes, the other part of my question was how do establish this convergence. The graph was supposed to be just an illustration. So assuming my variates are the y values of the function at different times, I would like to say that the distribution of the variate(s) at (or after) some time T is X. Is this a valid question to ask, and how would I answer it assuming I know the set of variates at least up to time T? – Peter R Oct 20 '16 at 17:56

0 Answers0