Unexpected TAD output

[graceannobrien]

lastdose outputs different TAD for records with same DATETIME (rows 22, 23, 24). I would expect all to have TAD ==0 since they occur at same time as dose (row 21), but only one record has TAD == 0.

Example data: temp_tad.csv

> temp_tad[21:24,]
   USUBJID            DATETIME EVID DVID DV ADDL II AMT      TAD     TAFD LDOS
21  hidden 2019-04-16 09:50:00    1    0 99    0  0   0  0.00000 982.9167    0
22  hidden 2019-04-16 09:50:00    0    5 99    0  0  NA  0.00000 982.9167    0
23  hidden 2019-04-16 09:50:00    0    6 99    0  0  NA 22.33333 982.9167    0
24  hidden 2019-04-16 09:50:00    0    4 99    0  0  NA 22.33333 982.9167    0

[andersone1]

@kylebaron Following up on this - it appears to be an issue, but we wanted to see what you thought.

[kylebaron]

Hey @andersone1 - sorry I didn't see this before. I need to check the code, but guessing that rows 23 and 24 are getting tagged as pre dose (they happen at the same time a the EVID 1 record, but they are getting sorted to happen before record 21 in sort order, so subject to the previous dose). When you sort on DATETIME, the actual order (record order) is random among records with the same value of DATETIME, so that's how it could happen. I thought I had some logic in the code to make sure to put all observations either before or after the dose at the same time to head this off, but will have to take a look. Thanks for reporting. Will look into this.

[andersone1]

@kylebaron

Our plan is to refrain from using lastdose::lastdose on new DAs until this issue is resolved. As a follow-up, is this a package you (or someone else) intend to maintain long-term?

[kylebaron]

Thanks, @andersone1 ; I'm working on a fix now and it should be ready to roll out first of the year. I'm expecting to continue to maintain the package.

[kylebaron]

I tagged @graceannobrien and @andersone1 to review this. It is able to handle the example data set @graceannobrien posted and another example while still hitting on existing tests. Please let me know if you have other test cases I could look at to make sure this is working as expected. Thanks!

[kylebaron]

@andersone1 - this update has been merged and is pending release. The current blocker is #49, which is waiting on approval of concept / behavior. @andersone1 @graceannobrien and @michaelmcd18 have been requested to review the OCC behavior in that PR.

[kylebaron]

Our plan is to refrain from using lastdose::lastdose on new DAs until this issue is resolved.

@andersone1 @graceannobrien @michaelmcd18 Reminder that this has been merged and can go out as soon as the OCC PR #49 is processed. Will push all of this out as immediately once that happens.

Thanks!