search

haowulab / SC2P

5 stars 4 forks source link

Error in simpleLoess(y, x, w, span, degree = degree, parametric = parametric, : span is too small #1

Open danshu opened 5 years ago

danshu commented 5 years ago

Hi,

I'm using SC2P and have the following error: Error in simpleLoess(y, x, w, span, degree = degree, parametric = parametric, : span is too small Calls: eset2Phase -> loess -> simpleLoess

Any suggestion on how to solving this problem?

Best, Danshu

haowulab commented 5 years ago

I'll have to see the data to figure this out. Can you reproduce this problem in a small portion of data?

dongyuanwu commented 4 years ago

Hello,

I met the same error when using the eset2Phase() function. And also got a warning message said: In eset2Phase(eset) : Zero proportion is greater than estimation. Any idea about them?

Thanks, Dongyuan

haowulab commented 4 years ago

It's a warning message, happened when estimated zero proportion is too high. This can be caused by the model assumption, eg, the data doesn't behave as our assumption. We developed the model on smart-seq data. It's possible that when read coverages are low, such as from 10X, this will happened. But we'll correct that and issue a warning. It's not an error and the downstream analysis should be okay.

hao

dongyuanwu commented 4 years ago

It's a warning message, happened when estimated zero proportion is too high. This can be caused by the model assumption, eg, the data doesn't behave as our assumption. We developed the model on smart-seq data. It's possible that when read coverages are low, such as from 10X, this will happened. But we'll correct that and issue a warning. It's not an error and the downstream analysis should be okay.

hao

Thanks so much for your quick reply! I was just wondering if it is the reason for the error message which I also met: Error in simpleLoess(y, x, w, span, degree = degree, parametric = parametric, : span is too small. This error occurred when I used eset2Phase function. Is there any solution now?

haowulab commented 4 years ago

As I said before, I'll have to see the data to figure this out. Can you reproduce the loess problem in a small portion of data?

dongyuanwu commented 4 years ago

As I said before, I'll have to see the data to figure this out. Can you reproduce the loess problem in a small portion of data?

I uploaded a demo for it. Hope it will work. https://drive.google.com/file/d/1zqmSiHF19OA_O4NMfM8EZa4pmRPFalla/view?usp=sharing