search

Teichlab / cellphonedb

MIT License
342 stars 105 forks source link

p-values of 0.0 #19

Closed HichamAffia closed 5 years ago

HichamAffia commented 5 years ago

Hi, I'm running CellPhoneDB with a --result-precision of 10 and I get a lot of straight 0.0 in the p-values output.

I wonder if I should consider them as p<0.0000000001 or if there is another way to look at it?

Like in the Fig5 a of the Single-cell reconstruction of the early maternal–fetal interface in humans paper I would like to plot the results for some clusters and to adjust the size of my dots according to the pvalue, so p-values of zero are a little problematic for me.

Thank you!

mief commented 5 years ago

Hi, Yes, that's also what we did to produce the dot plot, we replaced all zeros with very small numbers and then wrote >= certain -log10(p-value) in the legend. Best, Mirjana

rcw1 commented 5 years ago

I am having the same issue. I am using a single cell dataset with ~20k cells, running with the default 1000 permutations, and my p-values with CellPhoneDB are either 0.0 or 1.0. There are a few p-values that are in between, but not many. Is this what one should expect?

mief commented 5 years ago

Hi, Please uninstall and then install the new version. --result-precision is now by default 3.