Caveat in methods about bootstrap.

content/05.methods.md

@@ -56,6 +56,7 @@ After $N$ permutations (usually 10,000), we compute the $1 - p$ percentile of th
 > Using those resampled spectra and genotypes, we performed an aggregate mutation spectrum distance scan on the chromosome of interest and recorded the position of the marker with the largest adjusted cosine distance value.
 > We then defined a 90% confidence interval by finding two marker locations between which 90% of all $N$ bootstrap samples produced a peak cosine distance value. 
 > In other words, we estimated the bounds of the 90% confidence interval by finding the markers that defined the 5th and 95th percentiles of the distribution of maximum adjusted cosine distance values across $N$ bootstrap trials.
+> We note, however, that the bootstrap can exhibit poor performance in QTL mapping studies [@PMID:16783000]; namely, bootstrap confidence intervals tend to be larger than those estimated using either a "LOD drop" method or a Bayes credible interval, and can exhibit poorer-than-expected coverage (a measure of whether the confidence interval contains the true QTL location).
 
 #### Accounting for relatedness between strains