Regarding the question of estimating the error: I don't know a ton about stats, but can't you use bootstrapping to get a valid sense of the variability/margin o

Regarding the question of estimating the error: I don't know a ton about stats, but can't you use bootstrapping to get a valid sense of the variability/margin of error of your dataset for non normal distributions? http://en.wikipedia.org/wiki/Bootstrapping_(statistics) http://en.wikipedia.org/wiki/Bootstrapping_(statistics)