Implement occasion calculation by kylebaron · Pull Request #49 · metrumresearchgroup/lastdose

kylebaron · 2024-12-20T16:55:55Z

Optionally include occasion (OCC) in lastdose() (and friends) output. In general, a new occasion happens when there is a dose with "observation" records following before the next dose.

Status: work in progress; PR opened for review from the team. I'm working up these "Examples" so we can get a set of test cases that sufficiently cover the the range of behavior; this is tricky.

What we have so far:

OCC starts at 0; it doesn't increment until we have a dose with following observations
- I'd be open to starting OCC at 1 or maybe letting the user decide how to do it
When we say "observations", we mean EVID==0; EVID==2 don't count at this time
- We could include EVID==2 but IMO it makes it more difficult to define the occasion
- IMO, it should be rare to have a dose followed by EVID==2 record(s) followed by another dose with no EVID==0 in between
OCC increments at the dose record by looking ahead for following observation records; this was done intentionally

Questions:

Do you want OCC to show up automatically by default? or do you have to opt in?
- Currently you get it by default
What should OCC be for records prior to the first dose?
- Currently it is set to zero

How are BLQ observations handled?

If they are EVID==0 and uncommented, they will be considered as observations for OCC determination
If they are EVID==2 and uncommented, they will not be considered as observations for OCC determination
If they are commented, they will not be considered as observations for OCC determination
Users can "temporarily" comment these records (only during OCC calculation) through the comment argument

Some examples are included below and will form the test cases for the unit tests (not written yet).

Examples

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(mrgsolve)
#> 
#> Attaching package: 'mrgsolve'
#> The following object is masked from 'package:stats':
#> 
#>     filter
library(lastdose)

mod <- house(delta = 2, end = 24, outvars = "CP")

Single dose

dose <- evd(amt = 100)
data <- mrgsim_df(mod, dose, carry_out = "amt,evid,cmt")
data <- lastdose(data)

OCC starts at zero
OCC increments to 1 at the time of the first dose

data
#>    ID TIME evid amt cmt       CP TAD LDOS OCC
#> 1   1    0    0   0   0 0.000000   0    0   0
#> 2   1    0    1 100   1 0.000000   0  100   1
#> 3   1    2    0   0   0 4.247580   2  100   1
#> 4   1    4    0   0   0 4.228701   4  100   1
#> 5   1    6    0   0   0 3.861243   6  100   1
#> 6   1    8    0   0   0 3.496969   8  100   1
#> 7   1   10    0   0   0 3.164476  10  100   1
#> 8   1   12    0   0   0 2.863362  12  100   1
#> 9   1   14    0   0   0 2.590880  14  100   1
#> 10  1   16    0   0   0 2.344325  16  100   1
#> 11  1   18    0   0   0 2.121233  18  100   1
#> 12  1   20    0   0   0 1.919371  20  100   1
#> 13  1   22    0   0   0 1.736719  22  100   1
#> 14  1   24    0   0   0 1.571448  24  100   1


data <- mrgsim_df(mod, dose, carry_out = "amt,evid,cmt", recsort = 3)
data <- lastdose(data)

Here, OCC starts at 1 because it was the first record

data
#>    ID TIME evid amt cmt       CP TAD LDOS OCC
#> 1   1    0    1 100   1 0.000000   0  100   1
#> 2   1    0    0   0   0 0.000000   0  100   1
#> 3   1    2    0   0   0 4.247580   2  100   1
#> 4   1    4    0   0   0 4.228701   4  100   1
#> 5   1    6    0   0   0 3.861243   6  100   1
#> 6   1    8    0   0   0 3.496969   8  100   1
#> 7   1   10    0   0   0 3.164476  10  100   1
#> 8   1   12    0   0   0 2.863362  12  100   1
#> 9   1   14    0   0   0 2.590880  14  100   1
#> 10  1   16    0   0   0 2.344325  16  100   1
#> 11  1   18    0   0   0 2.121233  18  100   1
#> 12  1   20    0   0   0 1.919371  20  100   1
#> 13  1   22    0   0   0 1.736719  22  100   1
#> 14  1   24    0   0   0 1.571448  24  100   1

Multi-dose, with doses explicit in the data set

dose <- evd(amt = 100, ii = 12, addl = 1) %>% realize_addl()
data <- mrgsim_df(mod, dose, carry_out = "amt,evid,cmt")

OCC starts at 0 again, increments with the first dose
OCC increments at the time of the second dose because we have obserations following

lastdose(data)
#>    ID TIME evid amt cmt       CP TAD LDOS OCC
#> 1   1    0    0   0   0 0.000000   0    0   0
#> 2   1    0    1 100   1 0.000000   0  100   1
#> 3   1    2    0   0   0 4.247580   2  100   1
#> 4   1    4    0   0   0 4.228701   4  100   1
#> 5   1    6    0   0   0 3.861243   6  100   1
#> 6   1    8    0   0   0 3.496969   8  100   1
#> 7   1   10    0   0   0 3.164476  10  100   1
#> 8   1   12    0   0   0 2.863362  12  100   1
#> 9   1   12    1 100   1 2.863362   0  100   2
#> 10  1   14    0   0   0 6.838459   2  100   2
#> 11  1   16    0   0   0 6.573026   4  100   2
#> 12  1   18    0   0   0 5.982476   6  100   2
#> 13  1   20    0   0   0 5.416340   8  100   2
#> 14  1   22    0   0   0 4.901194  10  100   2
#> 15  1   24    0   0   0 4.434810  12  100   2

Multi-dose, with doses coded via addl

dose <- evd(amt = 100, ii = 12, addl = 1)
data <- mrgsim_df(mod, dose, carry_out = "amt,evid,cmt,addl,ii")

OCC increments at 12 hours, the time of the second dose

lastdose(data)
#>    ID TIME evid amt cmt ii addl       CP TAD LDOS OCC
#> 1   1    0    0   0   0  0    0 0.000000   0    0   0
#> 2   1    0    1 100   1 12    1 0.000000   0  100   1
#> 3   1    2    0   0   0  0    0 4.247580   2  100   1
#> 4   1    4    0   0   0  0    0 4.228701   4  100   1
#> 5   1    6    0   0   0  0    0 3.861243   6  100   1
#> 6   1    8    0   0   0  0    0 3.496969   8  100   1
#> 7   1   10    0   0   0  0    0 3.164476  10  100   1
#> 8   1   12    0   0   0  0    0 2.863362  12  100   2
#> 9   1   14    0   0   0  0    0 6.838459   2  100   2
#> 10  1   16    0   0   0  0    0 6.573026   4  100   2
#> 11  1   18    0   0   0  0    0 5.982476   6  100   2
#> 12  1   20    0   0   0  0    0 5.416340   8  100   2
#> 13  1   22    0   0   0  0    0 4.901194  10  100   2
#> 14  1   24    0   0   0  0    0 4.434810  12  100   2

Notice that we’ve (intentionally) made OCC to increment at 12 hours, even
though the TAD behavior is different with dose_first

lastdose(data, addl_ties = "dose_first")
#>    ID TIME evid amt cmt ii addl       CP TAD LDOS OCC
#> 1   1    0    0   0   0  0    0 0.000000   0    0   0
#> 2   1    0    1 100   1 12    1 0.000000   0  100   1
#> 3   1    2    0   0   0  0    0 4.247580   2  100   1
#> 4   1    4    0   0   0  0    0 4.228701   4  100   1
#> 5   1    6    0   0   0  0    0 3.861243   6  100   1
#> 6   1    8    0   0   0  0    0 3.496969   8  100   1
#> 7   1   10    0   0   0  0    0 3.164476  10  100   1
#> 8   1   12    0   0   0  0    0 2.863362   0  100   2
#> 9   1   14    0   0   0  0    0 6.838459   2  100   2
#> 10  1   16    0   0   0  0    0 6.573026   4  100   2
#> 11  1   18    0   0   0  0    0 5.982476   6  100   2
#> 12  1   20    0   0   0  0    0 5.416340   8  100   2
#> 13  1   22    0   0   0  0    0 4.901194  10  100   2
#> 14  1   24    0   0   0  0    0 4.434810  12  100   2

Multi-dose via addl, but no observations after the doses

dose <- evd(amt = 100, ii = 1, addl = 3)
data <- mrgsim_df(mod, dose, end = -1, add = c(0, seq(6,12)), carry_out = "amt,evid,cmt,addl,ii")

In this case, OCC doesn’t increment at the time of the dose at TIME==0;
This is because there wasn’t an observation after the first dose, so we
(intentionally) don’t increment OCC at that point

lastdose(data)
#>   ID TIME evid amt cmt ii addl       CP TAD LDOS OCC
#> 1  1    0    0   0   0  0    0  0.00000   0    0   0
#> 2  1    0    1 100   1  1    3  0.00000   0  100   0
#> 3  1    6    0   0   0  0    0 16.48841   3  100   1
#> 4  1    7    0   0   0  0    0 15.81578   4  100   1
#> 5  1    8    0   0   0  0    0 15.08405   5  100   1
#> 6  1    9    0   0   0  0    0 14.36032   6  100   1
#> 7  1   10    0   0   0  0    0 13.66355   7  100   1
#> 8  1   11    0   0   0  0    0 12.99826   8  100   1
#> 9  1   12    0   0   0  0    0 12.36465   9  100   1

We can see this explicitly here

dose <- evd(amt = 100, ii = 1, addl = 3) %>% realize_addl()
data <- mrgsim_df(mod, dose, end = -1, add = c(0, seq(6,12)), carry_out = "amt,evid,cmt,addl,ii")

The rule is: OCC doesn’t increment unless there are observations after the dose

data
#>    ID TIME evid amt cmt ii addl        CP
#> 1   1    0    0   0   0  0    0  0.000000
#> 2   1    0    1 100   1  0    0  0.000000
#> 3   1    1    1 100   1  0    0  3.391488
#> 4   1    2    1 100   1  0    0  7.639068
#> 5   1    3    1 100   1  0    0 11.987160
#> 6   1    6    0   0   0  0    0 16.488412
#> 7   1    7    0   0   0  0    0 15.815780
#> 8   1    8    0   0   0  0    0 15.084048
#> 9   1    9    0   0   0  0    0 14.360321
#> 10  1   10    0   0   0  0    0 13.663554
#> 11  1   11    0   0   0  0    0 12.998257
#> 12  1   12    0   0   0  0    0 12.364650
lastdose(data)
#>    ID TIME evid amt cmt ii addl        CP TAD LDOS OCC
#> 1   1    0    0   0   0  0    0  0.000000   0    0   0
#> 2   1    0    1 100   1  0    0  0.000000   0  100   0
#> 3   1    1    1 100   1  0    0  3.391488   0  100   0
#> 4   1    2    1 100   1  0    0  7.639068   0  100   0
#> 5   1    3    1 100   1  0    0 11.987160   0  100   1
#> 6   1    6    0   0   0  0    0 16.488412   3  100   1
#> 7   1    7    0   0   0  0    0 15.815780   4  100   1
#> 8   1    8    0   0   0  0    0 15.084048   5  100   1
#> 9   1    9    0   0   0  0    0 14.360321   6  100   1
#> 10  1   10    0   0   0  0    0 13.663554   7  100   1
#> 11  1   11    0   0   0  0    0 12.998257   8  100   1
#> 12  1   12    0   0   0  0    0 12.364650   9  100   1

EVID 2 or 3

These currently don’t count for establishing an occasion dose

data1 <- data.frame(
  ID = 1,
  TIME = c(0, 1, 2, 3, 4, 5, 6, 7),
  EVID = c(0, 1, 0, 1, 2, 1, 1, 0),
  AMT = 100
)

lastdose(data1)
#>   ID TIME EVID AMT TAD LDOS OCC
#> 1  1    0    0 100  -1    0   0
#> 2  1    1    1 100   0  100   1
#> 3  1    2    0 100   1  100   1
#> 4  1    3    1 100   0  100   1
#> 5  1    4    2 100   1  100   1
#> 6  1    5    1 100   0  100   1
#> 7  1    6    1 100   0  100   2
#> 8  1    7    0 100   1  100   2

data2 <- data.frame(
  ID = 2,
  TIME = c(0, 1, 2, 3, 4, 5, 6, 7),
  EVID = c(0, 1, 0, 1, 3, 1, 1, 0),
  AMT = 100
)

lastdose(data2)
#>   ID TIME EVID AMT TAD LDOS OCC
#> 1  2    0    0 100  -1    0   0
#> 2  2    1    1 100   0  100   1
#> 3  2    2    0 100   1  100   1
#> 4  2    3    1 100   0  100   1
#> 5  2    4    3 100   1  100   1
#> 6  2    5    1 100   0  100   1
#> 7  2    6    1 100   0  100   2
#> 8  2    7    0 100   1  100   2

But if we find an observation before the next dose, we start the occasion
there

data3 <- data.frame(
  ID = 3,
  TIME = c(0, 1, 2, 3, 4, 5, 6, 7),
  EVID = c(0, 1, 0, 1, 3, 0, 1, 0),
  AMT = 100
)

lastdose(data3)
#>   ID TIME EVID AMT TAD LDOS OCC
#> 1  3    0    0 100  -1    0   0
#> 2  3    1    1 100   0  100   1
#> 3  3    2    0 100   1  100   1
#> 4  3    3    1 100   0  100   2
#> 5  3    4    3 100   1  100   2
#> 6  3    5    0 100   2  100   2
#> 7  3    6    1 100   0  100   3
#> 8  3    7    0 100   1  100   3

Multiple subjects

data4 <- bind_rows(data1, data2, data3) %>% mutate(CMT = 1)
data4 <- mrgsim_df(mod, data4, end = -1, carry_out = "amt,evid,cmt,addl,ii")

lastdose(data4)
#>    ID TIME evid amt cmt ii addl        CP TAD LDOS OCC
#> 1   1    0    0   0   1  0    0  0.000000  -1    0   0
#> 2   1    1    1 100   1  0    0  0.000000   0  100   1
#> 3   1    2    0   0   1  0    0  3.391488   1  100   1
#> 4   1    3    1 100   1  0    0  4.247580   0  100   1
#> 5   1    4    2 100   1  0    0  7.739580   1  100   1
#> 6   1    5    1 100   1  0    0  8.476281   0  100   1
#> 7   1    6    1 100   1  0    0 11.789956   0  100   2
#> 8   1    7    0   0   1  0    0 15.729012   1  100   2
#> 9   2    0    0   0   1  0    0  0.000000  -1    0   0
#> 10  2    1    1 100   1  0    0  0.000000   0  100   1
#> 11  2    2    0   0   1  0    0  3.391488   1  100   1
#> 12  2    3    1 100   1  0    0  4.247580   0  100   1
#> 13  2    4    3 100   1  0    0  0.000000   1  100   1
#> 14  2    5    1 100   1  0    0  0.000000   0  100   1
#> 15  2    6    1 100   1  0    0  3.391488   0  100   2
#> 16  2    7    0   0   1  0    0  7.639068   1  100   2
#> 17  3    0    0   0   1  0    0  0.000000  -1    0   0
#> 18  3    1    1 100   1  0    0  0.000000   0  100   1
#> 19  3    2    0   0   1  0    0  3.391488   1  100   1
#> 20  3    3    1 100   1  0    0  4.247580   0  100   2
#> 21  3    4    3 100   1  0    0  0.000000   1  100   2
#> 22  3    5    0   0   1  0    0  0.000000   2  100   2
#> 23  3    6    1 100   1  0    0  0.000000   0  100   3
#> 24  3    7    0   0   1  0    0  3.391488   1  100   3

^{Created on 2025-01-02 with reprex v2.1.1}

kylebaron · 2025-01-08T18:42:47Z

Check behavior for commented doses.

The previous two commits covered most of the data.frame() calls. Adjust two more spots to resolve the remaining test failures under R 3.6.

kylebaron · 2025-01-22T13:40:19Z

Thanks for help with stringsAsFactors again, @kyleam.

kylebaron · 2025-01-28T13:01:01Z

Waiting on @andersone1 @graceannobrien @michaelmcd18 to review behavior and either confirm or propose alternate behavior.

kylebaron · 2025-02-21T12:42:32Z

@kyleam This is ready for review; I'd be happy to talk through any of the C++ code.

kyleam

Some of the conceptual details you laid out in the PR description are a bit beyond me, but it looks like you did a very thorough job of documenting those in inst/test-data/occ/occ-data.R, the lastdose docstring, and tests/testthat/test-occ.R.

Aside from my few comments inline, this looks good to me code-wise (and inline with what I expect from inspecting the surrounding code).

inst/test-data/occ/occ-data.R

R/lastdose.R

tests/testthat/test-occ.R

kyleam

Updates look good to me, thanks.

(I know my comment about colliding test descriptions is still unresolved, but that's not blocking.)

kylebaron · 2025-02-25T16:18:40Z

All of @kyleam review comments have been resolved (I believe); only waiting on input from Metrum datascience (@andersone1 , @michaelmcd18 , @graceannobrien ) on behavior.

kylebaron added 6 commits June 3, 2024 22:06

bump development version

9cd08a3

working on occ

fa1519a

fix conflicts

1d8dd44

implement occ calculation

7ec02ce

optional inclusion of occ

57891ef

tweak test

a10217c

kylebaron marked this pull request as draft December 20, 2024 17:18

kylebaron added 2 commits December 20, 2024 17:11

tweak occ logic

31244ba

add exclude to Rbuildignore

bc47cd8

kylebaron requested review from andersone1, graceannobrien and michaelmcd18 December 21, 2024 01:57

kylebaron changed the title ~~Implement Occasion calculation~~ Implement occasion calculation Dec 21, 2024

kylebaron added 5 commits December 30, 2024 19:28

add test data; start working through tests

4471f7a

tests

e9cf2e0

rename data directory

4d05903

rename test daata directory

8608e81

add tests for multiple subjects

e9f72eb

kylebaron mentioned this pull request Jan 6, 2025

Add OCC to lastdose output #50

Closed

kylebaron linked an issue Jan 6, 2025 that may be closed by this pull request

Add OCC to lastdose output #50

Closed

kylebaron and others added 5 commits January 9, 2025 08:39

handle commented records

3259225

Merge branch 'main' into occ

eab12da

tweak read.csv

cf242a2

strings as factors globally

cdc8e5a

tests: specify missing stringsAsFactors for R 3.6 compatibility

8a2f8e5

The previous two commits covered most of the data.frame() calls. Adjust two more spots to resolve the remaining test failures under R 3.6.

kylebaron mentioned this pull request Jan 28, 2025

Unexpected TAD output #46

Open

kylebaron marked this pull request as ready for review February 12, 2025 13:02

kylebaron requested a review from kyleam February 21, 2025 12:41

documentation

4e7cdd1

kyleam requested changes Feb 21, 2025

View reviewed changes

inst/test-data/occ/occ-data.R Outdated Show resolved Hide resolved

R/lastdose.R Outdated Show resolved Hide resolved

R/lastdose.R Outdated Show resolved Hide resolved

tests/testthat/test-occ.R Outdated Show resolved Hide resolved

kylebaron added 3 commits February 25, 2025 07:40

re-name directory

829474d

remove inaccurate documentation

e85c670

pass include_occ along to prevent unneeded work

ed79f6e

kyleam approved these changes Feb 25, 2025

View reviewed changes

clarify test names

af2565b

kylebaron merged commit c66be6b into main Mar 21, 2025
7 checks passed

kylebaron deleted the occ branch March 21, 2025 11:19

kylebaron mentioned this pull request Mar 21, 2025

Release 0.5.0 #52

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comments

Implement occasion calculation#49

Implement occasion calculation#49
kylebaron merged 23 commits intomainfrom
occ

kylebaron commented Dec 20, 2024 •

edited

Loading

Uh oh!

kylebaron commented Jan 8, 2025

Uh oh!

kylebaron commented Jan 22, 2025

Uh oh!

kylebaron commented Jan 28, 2025

Uh oh!

kylebaron commented Feb 21, 2025

Uh oh!

kyleam left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kyleam left a comment

Uh oh!

kylebaron commented Feb 25, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Comments

Conversation

kylebaron commented Dec 20, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Examples

Single dose

Multi-dose, with doses explicit in the data set

Multi-dose, with doses coded via addl

Multi-dose via addl, but no observations after the doses

EVID 2 or 3

Multiple subjects

Uh oh!

kylebaron commented Jan 8, 2025

Uh oh!

kylebaron commented Jan 22, 2025

Uh oh!

kylebaron commented Jan 28, 2025

Uh oh!

kylebaron commented Feb 21, 2025

Uh oh!

kyleam left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kyleam left a comment

Choose a reason for hiding this comment

Uh oh!

kylebaron commented Feb 25, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kylebaron commented Dec 20, 2024 •

edited

Loading