Factor value that is part of `keep` is converted to numeric in the source data

[billdenney]

d_sim$TRTG should not be modified, but in the first pass where it is part of keep, it is. (My guess is that this will end up being an rxode2 issue, but I'm not sure exactly where keep has its effects at the moment, and I want to put the reprex into the record.)

library(nlmixr2est)
#> Loading required package: nlmixr2data

one.compartment <- function() {
  ini({
    tka <- 0.45; label("Ka")
    tcl <- 1; label("Cl")
    tv <- 3.45; label("V")
    eta.ka ~ 0.6
    eta.cl ~ 0.3
    eta.v ~ 0.1
    add.sd <- 0.7
  })
  model({
    ka <- exp(tka + eta.ka)
    cl <- exp(tcl + eta.cl)
    v <- exp(tv + eta.v)
    d/dt(depot) = -ka * depot
    d/dt(center) = ka * depot - cl / v * center
    cp = center / v
    cp ~ add(add.sd)
  })
}

d_sim <- theo_sd
d_sim$TRTG <- ordered(ifelse(theo_sd$WT < median(theo_sd$WT), "Low", "High"))
d_sim$TRTG
#>   [1] High High High High High High High High High High High High High High High
#>  [16] High High High High High High High High High High High High High High High
#>  [31] High High High High High High High High High High High High High High High
#>  [46] High High High Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#>  [61] High High High High High High High High High High High High Low  Low  Low 
#>  [76] Low  Low  Low  Low  Low  Low  Low  Low  Low  High High High High High High
#>  [91] High High High High High High High High High High High High High High High
#> [106] High High High Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> [121] Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> [136] Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> Levels: High < Low
fit <- nlmixr2(one.compartment, d_sim, est="rxSolve", list(keep = "TRTG"))
d_sim$TRTG
#>   [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
#>  [38] 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 2 2
#>  [75] 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2
#> [112] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

d_sim <- theo_sd
d_sim$TRTG <- ordered(ifelse(theo_sd$WT < median(theo_sd$WT), "Low", "High"))
fit <- nlmixr2(one.compartment, d_sim, est="rxSolve")
d_sim$TRTG
#>   [1] High High High High High High High High High High High High High High High
#>  [16] High High High High High High High High High High High High High High High
#>  [31] High High High High High High High High High High High High High High High
#>  [46] High High High Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#>  [61] High High High High High High High High High High High High Low  Low  Low 
#>  [76] Low  Low  Low  Low  Low  Low  Low  Low  Low  High High High High High High
#>  [91] High High High High High High High High High High High High High High High
#> [106] High High High Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> [121] Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> [136] Low  Low  Low  Low  Low  Low  Low  Low  Low 
#> Levels: High < Low

^{Created on 2023-09-25 with reprex v2.0.2}

[billdenney]

With a quick additional test, it seems like this only has an effect when nStud = 1 (the default) and not when nStud = 1000.

[mattfidler]

d_sim$TRTG will be modified since there are dosing records that will be inserted and interpolations that will be performed.

But there should be no effects on the source data.

[mattfidler]

library(rxode2)
#> rxode2 2.0.14.9000 using 8 threads (see ?getRxThreads)
#>   no cache: create with `rxCreateCache()`

one.compartment <- function() {
ini({
    tka <- 0.45; label("Ka")
    tcl <- 1; label("Cl")
    tv <- 3.45; label("V")
    eta.ka ~ 0.6
    eta.cl ~ 0.3
    eta.v ~ 0.1
    add.sd <- 0.7
    })
model({
      ka <- exp(tka + eta.ka)
      cl <- exp(tcl + eta.cl)
      v <- exp(tv + eta.v)
      d/dt(depot) = -ka * depot
      d/dt(center) = ka * depot - cl / v * center
      cp = center / v
      cp ~ add(add.sd)
      })
}

d_sim <- nlmixr2data::theo_sd
d_sim$TRTG <- ordered(ifelse(nlmixr2data::theo_sd$WT < median(nlmixr2data::theo_sd$WT), "Low", "High"))
summary(d_sim$TRTG)
#> High  Low 
#>   84   60

fit <- rxSolve(one.compartment, d_sim, keep="TRTG")

summary(d_sim$TRTG)
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>   1.000   1.000   1.000   1.417   2.000   2.000

^{Created on 2023-12-04 with reprex v2.0.2}

[mattfidler]

This is because your factor is an ordered factor. I'm unsure if I will allow them to be kept or not. I can't support every different type of data since it actually isn't kept and merged at the end.

[mattfidler]

We already tried an approach where we merged back the data and it didn't successfully run (I remember you tried it).

[mattfidler]

So currently in keep:

• Standard factors are supported
• Standard logical vectors are supported
• Standard character vectors are supported
• Standard integer vectors are supported

Any sort of attributes will be dropped (maybe should have some sort of warning but would need to explicitly check for attributes).

Could additionally support ordered factors