But given an independent measure of time of divergence, you should be able
to define an upper limit and confidence interval for the rate of change,
even with zero events, no? The difference between zero events and a
single event in determining a rate is very small, even though the latter
yields a defined value and the former does not. And if there are events
in at least one of the two categories, as you suggest, you should be able
to determine if there is a significant difference.
On Wed, 16 Oct 2002, Aaron J Mackey wrote:
> If two sequences are 100% identical (or, at least at synonymous sites),
> then there's nothing you can say about the rate of synonymous
> substitution, it's an unknown. It's like dividing by zero, no matter how
> you might think that you understand it to be "Infinity", it's just
> undefined, full stop.
>> One method is to prune your tree of one of the "duplicate" sequences. How
> many nonsynonymous sites are there? One or two? Many? I guess if there
> are many, then you'd want to pursue this further in order to try to
> convince yourself of positive selection, but if there's only a few then I
> don't think you can say much and should just move on to other more
> interesting comparisons ...