search

satijalab / sctransform

R package for modeling single cell UMI expression data using regularized negative binomial regression
GNU General Public License v3.0
207 stars 33 forks source link

Do the residuals follow any distribution? #23

Closed mochar closed 5 years ago

mochar commented 5 years ago

If not, are the deviance residuals any worse than the pearson residuals? I think they are normally distrubited right?

ChristophH commented 5 years ago

You can look at both residual types and evaluate their distributions. The vst function accepts a residual_type parameter which can be either pearson or deviance (pearson by default)

For most genes the distribution of pearson residuals may show a long tail, but they have the nice property of mean 0 and variance 1 for non-differentially expressed genes. This is not true of the deviance residuals.

mochar commented 5 years ago

Thank you for the awesome response!