Advanced usage

## DO NOT MODIFY THIS BLOCK (unless you know what you're doing)
library(knitr)
library(rprojroot)
opts_knit$set(root.dir = find_root(has_file(".Rprofile") | 
                                     is_rstudio_project | 
                                     is_r_package | 
                                     is_git_root))
opts_chunk$set(echo = TRUE)
opts_chunk$set(message = TRUE)
## LOAD PACKAGES HERE
library(NMproject)
library(dplyr)
m1 <- readRDS("Results/m1.RDS")

Setting initial estimates

You've already seen how setting an init vector allows setting of values, We can also give named vectors like so:

m1 <- m1 %>% init_theta(init = c(KA = 1), 
                        FIX = c(KA = TRUE))

m1 %>% dollar("THETA")
## return a tibble version of $OMEGA with init_omega()
io <- m1 %>% init_omega()
io 

# this can then be manipulated (note that io is a list of a data.frame) and put
# back into the object m1 with: 

# m1 <- m1 %>% init_omega(io) 

# see next chunk
## here we will manipulate io with the ready made block() function
## this constructs
io <- io %>% block(c(2,3))  ## make block out ETA 2 and 3
## put modified io wit
m1 <- m1 %>% init_omega(io)
m1 %>% dollar("OMEGA")
## for demo purposes we'll reverse the process with unblock()
io <- m1 %>% init_omega()
io <- io %>% unblock(c(2,3))
m1 <- m1 %>% init_omega(io)
m1 %>% dollar("OMEGA")

Perturbing initial parameters

To modify initial estimates, we'll use the mutate-like behaviour of init_* functions. We will modify the init by referencing itself. We'll modified all our fixed effects (log transformed) by 30%

m1 %>% dollar("THETA")
m1 <- m1 %>% init_theta(init = rnorm(init, mean = init, sd = 0.3))
m1 %>% dollar("THETA")

As a more advanced exercise, lets create 5 runs with modified initial estimates by combining the above with dplyr::mutate()

dp <- tibble(rep = 1:5) %>%
  mutate(
    m = m1 %>% ## start with single parent run, m1, an nm object of length 1
      child(run_id = rep) %>% ## rep is vector of length 5
      ## this makes the child() output a nm object vector of length 5 - each with a unique run_id
      ## do the same manipulation above - since NMproject functiosn are
      ## vectorized, this will apply the same manipulation elementwise
      init_theta(init = rnorm(init, mean = init, sd = 0.3))
  )

## the m column of dp is a nm object vector, display all $THETAs like this
dp$m %>% dollar("THETA")

## run this with 
# dp$m %>% run_nm()

Automating model checking

We'll use a similar combination of nm object vectorized and dplyr the previous section to test multiple model structures in one go using the subroutine() control file manipulation function. .available_advans shows all available advans.

## discard previous changes to m1 by reloading
m1 <- readRDS("Results/m1.RDS")

.available_advans ## display available advans

dt <- .available_advans %>%
  ## filter only for oral dosing advans
  filter(oral %in% TRUE) %>%
  ## mutate state create a column vector m of nm objects
  ## first step is to create children runs from the parent object m1
  ## this is done by supplying a vector of run_ids to the child() function
  mutate(m = m1 %>% ## start with parent m1 again
           child(run_id = label) %>% ## create multiple children using "label" column
           ## then use advan and trans columns with subroutine()
           subroutine(advan = advan, trans = trans))

## view the $PK blocks of each
dt$m %>% dollar("PK")

## run them all and wait for them to finish
dt$m %>% run_nm() %>% wait_finish()

summary_wide(dt$m)

#dt$m %>% 
#  subset(ofv(.) < 120) %>%
#  nm_render("Scripts/basic_gof.Rmd")

Often you'll want to know the right level of parallelisation to run your model to maximise speed without wasting too many resources. The following uses the vectorized nm objects in a similar way with dplyr::mutate() to create multiple runes with different levels of parallelisation. For the demo we'll just test it across 1, 2 and 3 cores, but this can be an arbitrary vector.

dc <- tibble::tibble(cores = c(1, 2, 3)) %>%  ## only 3 different values to test for demo purposes
  mutate(m = m1 %>% ## start with parent m1 again
           ## supply the "cores" vector (of the tibble) to child(), this will create 3 runs
           child(run_id = cores) %>%
           ## run them all in m1_coretest
           run_in("Models/m1_coretest") %>%
           ## run them all with the following execute command
           cmd(parallel_execute) %>%
           ## and the following parafile
           parafile("/opt/NONMEM/nm75/run/mpilinux8.pnm") %>%
           ## and finally set the cores to the "cores" vector
           cores(cores))

## run them all and wait for them to finish
dc$m %>% run_nm() %>% wait_finish()

## following is disabled for the demo
# job_info(dc$m)

## plot cores vs Rtime or Ttime to get plots of run time and total time vs number of CPUs
## This

You can also use the vectorized nature to create a simulation re-estimation routines from scratch.

## whenever making a routine that will scale up to 100s or 1000s or runs, always start with 1 or 2 replicates.  Get the code working below, and then rerun with desired number of replicates (here we're just using 3 for demo purposes).  You will not waste time because run_nm() will skip over runs that have already completed. 
n_sims <- 3  

dsr <- tibble(sim = 1:n_sims) %>%
  mutate(
    msim = m1 %>%              ## start with single parent run, m1, an nm object of length 1
      update_parameters() %>%  ## update inits to finals, here nm object is still length 1
      child(run_id = sim) %>%  ## at this point it becomes length n_sims
      run_in("Models/m1_simest") %>% ## this applies the run_in modification to all n_sims runs
      convert_to_simulation(subpr = 1, seed = sim) ## converts all to simulation
  )

## run, wait, read results and then write to run_dir paths of simulations
dsr$msim %>% run_nm() %>% 
  wait_finish() %>%
  output_table(only_append = "DV_OUT") %>%
  write_derived_data(file.path(run_dir_path(dsr$msim), "simdata.csv"))

## Now create mest column
dsr <- dsr %>%
  mutate(
    mest = m1 %>% child(run_id = sim) %>%  ## estimations derived from m1
      run_in("Models/m1_simest/est") %>%   ## run in a new subdirectory
      data_path(file.path(run_dir_path(msim), "simdata.csv")) %>%    ## set new data_paths
      ## refill $INPUT. Rename DV to be DV_OUT column. Run nm_diff() command below to see
      ## what has changed
      fill_input(rename = list("DV_OBS" = "DV", "DV" = "DV_OUT"))
  )

# nm_diff(dsr$mest[1])

dsr$mest %>% run_nm() %>% 
  wait_finish() %>%
  summary_wide(parameters = "all")


Try the NMproject package in your browser

Any scripts or data that you put into this service are public.

NMproject documentation built on Sept. 30, 2022, 1:06 a.m.