scale parameter does not apply to endpoint

[danieleboaretti]

library(nlmixr2)

one.compartment <- function() {
  ini({
    tka <- log(1.57); label("Ka")
    tcl <- log(2.72); label("Cl")
    tv <- log(31.5); label("V")
    eta.ka ~ 0.6
    eta.cl ~ 0.3
    eta.v ~ 0.1
    add.sd <- 0.7
  })
  # and a model block with the error specification and model specification
  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)
  })
}
scale <- 1
aa <- rxSolve(one.compartment, seed=1, events=theo_sd,
  scale=c(center=scale, cp=scale))
head(aa %>% select(time, center, cp))

scale <- 0.1
aa <- rxSolve(one.compartment, seed=1, events=theo_sd,
  scale=c(center=scale, cp=scale))
head(aa %>% select(time, center, cp))

Hello, I noticed that the scale parameter does not apply to the endpoint cp.
This is the output with scale=1:

  time   center       cp
1 0.00   0.0000 0.000000
2 0.25 128.9277 1.933501
3 0.57 216.7502 3.250556
4 1.12 271.6130 4.073321
5 2.02 276.7748 4.150732
6 3.82 242.2918 3.633598

This is the output with scale=0.1:

  time   center       cp
1 0.00    0.000 0.000000
2 0.25 1289.277 1.933501
3 0.57 2167.502 3.250556
4 1.12 2716.130 4.073321
5 2.02 2767.748 4.150732
6 3.82 2422.918 3.633598

is it intended? I think it would be useful to include the endpoint if possible.

[mattfidler]

The scale was applied to the "classic" RxODE models, I am have to look into the code to see why it doesn't work.

[danieleboaretti]

I think the scaling should be introduced for the LHS that are not states because at the moment the scaling applies to all states including states that are endpoint.
I think the addition for the LHS should be done here:
https://github.com/nlmixr2/rxode2/blob/main/src/rxode2_df.cpp#L761
it might need to have an extra argument to rxSolve to iterate over the right index for scaling there

[mattfidler]

I am trying to make nlmixr2 and rxode2 syntax and models the same; The scaling is important for nlmixr2 and rxode2, so I'm unsure if the old solution is the right solution.

I will keep this in mind. I might fix this in the future, though this is lower on my priority list.

Currently, you can do scaling after solving or before with model piping.

[mattfidler]

As a note, to estimate in focei and other such algorithms, the scaling needs to be in the model itself.