Application of mashr to smaller number of tests

[vinettey]

Hi! Thanks for the wonderful tool that your team developed! I was wondering if this method could be applied to scenarios outside eQTL where the matrix of effects is smaller (e.g. 1000 rows/tests in 10 conditions). Is there a lower limit of number of tests included for the algorithm to learn the significant and random effects from the data? Is there also a requirement for how sparse the data should be? Thank you!

[pcarbo]

@vinettey In principle, mashr should work for your data set, but yes in general it is more challenging to analyze data sets with fewer samples and/or more c onditions. How to analyze data sets with more conditions and/or fewer samples was the focus of a recent paper we wrote; in particular, we gives some suggestions for how to more accurately estimate the prior covariance matrices. Hope this helps.