Error on getPsModel about cohortmethod

[schuemie]

My first guess is that results is not a valid object as generated by createPs, suggesting the problem already occurred earlier. Could you try

str(attr(results, "coefficients"))

and see if it confirms it is a Named num? If not, can you see if createPs threw an error?

[fdefalco]

str(attr(results,"coefficients"))
Named num [1:1547] -1.19 0 0 0 0 ...

• attr(*, "names")= chr [1:1547] "(Intercept)" "17" "18" "19" ...

It appears to be a Named num, createPs did not throw any errors.

[schuemie]

Ah, I'm guessing all coefficients except the intercept are 0. Are you by any chance using a really small example data set?

[fdefalco]

Yes, that does appear to be the case when I look at the data coming back from attr(results,"coefficients"). I'm not sure what is considered small, but in this case I have 11k people in my target cohort, 38k in my comparator and 3k in the outcome cohort. I'll email you the R code so you can check it out first hand.

[schuemie]

We figured out that this was caused by not excluding the target and comparator drugs from the covariates. The new version of CohortMethod will mention the need for this more clearly.