prm_logit_normal: Parameter with logit-normal distribution
In assemblerr: Assembly of Pharmacometric Models

Description Usage Arguments Details Value See Also Examples

This building block declares a parameter model for a parameter that follows the normal distribution on the logit-scale.

1	prm_logit_normal(name, mean_logit = 0, var_logit = 1)

`name`	Parameter name
`mean_logit`	Mean on the logit scale
`var_logit`	Variance on the logit scale

Parameter models specify type, name, and values for a parameter. The parameter model type is selected through the function name. The parameter name and values are provided as function arguments.

Parameter names

Every parameter must have a valid name. A parameter name can contain letters, numbers as well as the underscore character. The name needs to start with a letter.

Adding a parameter with an already existing name will replace the definition of the parameter. For example, the parameter “base” will have a log-normal distribution in the following snippet:

1
2
3

m <- model() +
 prm_normal("base") +
 prm_log_normal("base")

Parameter values

The parameter values that a parameter model expects vary by type. For example, prm_normal() requires the mean and the variance, whereas for prm_log_normal() median and variance on the log scale need to be provided. The argument name should indicate what parameter value is expected.

MU-referencing

assemblerr can include mu-referencing statements for parameter distributions that support it. The functionality can be activated by setting the option prm.use_mu_referencing to TRUE as shown in the following snippet:

m <- model() +
  prm_normal("base") +
  prm_log_normal("slp") +
  obs_additive(response~base+slp*time)

render(
  model = m,
  options = assemblerr_options(prm.use_mu_referencing = TRUE)
)

A building block of type 'parameter'

Other parameter models: prm_log_normal(), prm_no_var(), prm_normal()

# EMAX dose-response model with emax (log-normal) and ed50 (no variability) parameters
m2 <- model() +
  input_variable("dose") +
  prm_log_normal("emax", 10, 0.3) +
  prm_no_var("ed50", 5) +
  obs_proportional(effect~emax*dose/(ed50+dose))

# a log-normal parameter that is directly observed
m <- model() +
  prm_log_normal("wt") +
  obs_additive(~wt)