R help - Feb 2005 - off-topic: bootstrapping ratios of variances to get CIs (or alternative approaches)

Hi,
 
Apologies, this is a stats question rather than an R question. I am trying to
obtain reliable confidence limits for a phenotypic measure of QST. I'm
calculating this as var(between populations) / (var(between popns)+2*var(within
popns)) using nlme. I have tried bootstrapping, using diffferent methods to
calculate the confidence intervals, however the type one errors of this approach
are much greater than the alpha value of 0.05. My problem is that the number of
populations I have is small (only 9) and I believe non-parametric bootstrapping
performs poorly when estimating confidence limits for variances when the sample
size is small.
 
I wondered if anyone could suggest a method to correct for this or an approach
other than bootstrapping that may yield reliable 95% confidence limits.
 
Many thanks,
Albert

	[[alternative HTML version deleted]]