epzjlm
commented
8 years ago Hi Rene,
Yes Gib and I discussed this before. In ARIES, there are zero's but not all of them. So I think we should either remove eosinophils for every study or only adjust if you have more than xx % of non zero values. For some of the case control studies such as asthma we need adjustment though.
Hi there! In 04a-methylation_variables.sh cell type proportions are estimated, after which the different numeric covariates (including those cell type proportions) are checked for normality for men and women separately, and then transformed. The eosinophils are on the low end of the spectrum. In the Leiden data all men and most women have estimated values of exactly zero. Not only does this seem highly unlikely, it also leads to errors when testing for normality (shapiro.test) and later when fitting linear models with the proportions as an outcome. Has anyone else observed such low values for eosinophils?
On a more practical note, I think these estimates need to be checked. Or at least the error should be caught, e.g. using try(), which is an easy fix.