Open lucia-ramirez opened 3 years ago
Hi Lucia,
Thank you for your question.
I’m not sure whether I understand your question correctly. Do you mean you want to test between two models like y=A+B+C+D and y=A+B?
This kind of testing is not implemented in the current version, but I’ll consider adding this possibility in the next version.
Best regards,
Liang
From: Lucia Ramirez Navarro notifications@github.com Sent: Tuesday, March 2, 2021 2:56 PM To: lhe17/nebula nebula@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [lhe17/nebula] Testing coefficients (#1)
Is it possible to test for multiple coefficients of the model rather than having a the pvalue and statistics for each coefficient? Thanks
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/1 , or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUWNJCUJ7LVVPWJRBJTTBU7FFANCNFSM4YPUMJDQ . https://github.com/notifications/beacon/AGDISUWRYB2XBG5NQD23PRTTBU7FFA5CNFSM4YPUMJD2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4MHFLGLQ.gif
Hi Liang,
Yes, I am taking about two scenarios: the first one is the one you mention where you test if including an additional covariate improves the model and the second one, is for example when a covariate has multiple categories so instead of having a pvalue for each one, you want the overall effect. Something analogous to what contrasts.fit from limma does. This last thing could easily by implemented with glht from multcomp, the only thing is that it needs the complete model statistics rather than the summary of it that it is returned from nebula
Thanks,
Lucia
Hi Lucia,
OK. I see your point now. Thank you for raising this question.
Yes, you are right. This kind of test can be done by e.g., a likelihood ratio test, which requires additional statistics present in the output. I’ll consider making this kind of test available in the following updates.
Best regards,
Liang
From: Lucia Ramirez Navarro notifications@github.com Sent: Wednesday, March 3, 2021 6:31 AM To: lhe17/nebula nebula@noreply.github.com Cc: lhe17 hyx520101@gmail.com; Comment comment@noreply.github.com Subject: Re: [lhe17/nebula] Testing coefficients (#1)
Hi Liang,
Yes, I am taking about two scenarios: the first one is the one you mention where you test if including an additional covariate improves the model and the second one, is for example when a covariate has multiple categories so instead of having a pvalue for each one, you want the overall effect. Something analogous to what contrasts.fit from limma does. This last thing could easily by implemented with glht from multcomp, the only thing is that it needs the complete model statistics rather than the summary of it that it is returned from nebula
Thanks,
Lucia
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/1#issuecomment-789648364 , or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUSAZW2C7K53TEC2WTTTBYMYTANCNFSM4YPUMJDQ . https://github.com/notifications/beacon/AGDISUTKHW3YT7EINA2MCETTBYMYTA5CNFSM4YPUMJD2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOF4IRH3A.gif
Is it possible to test for multiple coefficients of the model rather than having a the pvalue and statistics for each coefficient? Thanks