-
Notifications
You must be signed in to change notification settings - Fork 31
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
qdrg with MICE (as.mira) - is there any way to marry them? #446
Comments
OK, I think I found a quick-and-(very)dirty workaround. I looked at the source in this file: Line 139 in 3957a59
namely to this function:
This creates the emm.basis, which is kinda "input" for the emmeans. It prepares the necessary components, like pooled coefficients and (co)variances, passed later to emmeans. And that's exactly what I need to pass to qdrg: coefficients, var_cov and df. So I took your code to pool the estimates:
I plotted values of selected statistics for each imputed dataset and the pooled ones. The consistency is perfect, so it worked well. So now I have a universal way of dealing with unknown models in the framework of multiple imputation. At least as long, as all necessary components will be provided... Would you consider adding kind of a "helper function" that facilitates these steps in case someone is using the qdrg()? It could help a lot of people! (I guess...) |
Thanks. I don't want to export that function as it has such a narrow purpose. However, I am providing a link to this issue in Section I of the |
By the way, |
Thank you. Yes, you are right. Anyway, it's great the qdrg() facilitates "bridging" emmeans to models in so many situations. It was a great idea to make it. |
I need to run ordinal logistic regression for repeated observations via GEE.
I can use either ordLORgee from the multgee package or repolr from the repolr package.
Unfortunately, neither supports emmeans. They can be integrated with emmeans via qdrg, though. This works very well.
Now, I have multiply imputed dataset. I know that emmeans works with via as.mira() - I use it with mmrm and gee - but this seems to work with models that emmeans can recognize.
But how to deal with qdrg this way? I'm afraid it's impossible, because how to pass all those coefficients and variances?
For example (from another question):
but what to do with qdrg?
I thought, that maybe I should create qdrg in each iterated analysis:
and somehow pass it to emmeans, but it also doesn't work:
I desperately try to complete the analysis... I'm about to write the components necessary to implement the emmeans through the vignette, but I thought there is some workaround that can save me this time?
The text was updated successfully, but these errors were encountered: