OHDSI Home | Forums | Wiki | Github

Estimation error: what is the mean of "Currently only MDRR for Cox is supported" when using atlas to do estimation?


(panpan) #1

dear all,
when I use the atlas to do estimation, the log in the part of diagnosis shows that: Thread 1 returns error: “Currently only MDRR for Cox is supported”, like the picture.
what are the mean of the error and the following errors? and how can I solve it?
thank you very much


(Martijn Schuemie) #2

Yes, sorry about that. The main use of estimation until now has been to use time-to-event (Cox proportional) hazards models, and some of the code is specific to those types of models.

Specifically, here the problem is that one of the diagnostics is to compute the statistical power (expressed as minimum detectable relative risk). The CohortMethod function computeMdrr currently only implements power calculations for Cox models, but apparently you’re using logistic regression or Poisson instead.

For now, the solution would be to just remove this code from your estimation package, so it doesn’t try to compute the MDRR.

We should fix this overall (looping in @jweave17 and @anthonysena ), for now by skipping power calculations if the outcome model isn’t a Cox model. In the future, we’d want to implement power calculations for the other outcome models.


(Martijn Schuemie) #3

Created an issue


(panpan) #4

thank you very much @schuemie, as you say ,it is because of the model. I choose the COX model then the error disappear. but there is still a problem like the picture.
what is the wrong? and does it has any effenct for the estimation?


(Martijn Schuemie) #5

I have no idea. I’ve never seen this message before. It seems to be uninstalling a package automatically. I also haven’t seen ‘Applying project Rprofile’ before. Are you using a profiler?


t