Observation models for continuous data

Purpose

The observation model is the link between the prediction f of the structural model and the observation. Thus, the observational model is an error model representing the noise and the uncertainty of the measurements. The observation model can be defined in the Mlxtran model file and/or in the Monolix interface. If the observation model is defined in the Mlxtran model file, then the error model settings in the Monolix interface are disabled.

Possible observation models

For the continuous observations, the general form y = f + g e is considered where e is a sequence of independent random variables normally distributed with mean 0 and variance 1. Some extensions assume that there is a distribution t such that t(y) = t(f) + g e. It is also possible to assume that the residual errors are correlated. Following is a list of observational models that can be selected in the Monolix user interface:

  • const: constant error model y = f + ae
  • prop: proportional error model y = f + bfe
  • comb1: combined error model y = f + (a+bf)e
  • comb2: combined error model y = f + \sqrt{a^2 + (bf)^2}e (equivalent to y = f + ae_1 + bfe_2 where e1 and e2 are sequences of independent random variables normally distributed with mean 0 and variance 1)
  • propc: proportional error model + power y = f + bf^ce
  • comb1c: combined error model + power y = f + (a + b f^c)e
  • comb2c: combined error model + power y = f + \sqrt{a^2 + (bf^c)^2}e (equivalent to y = f + ae_1 + bfe_2 where e1 and e2 are sequences of independent random variables normally distributed with mean 0 and variance 1)
  • exp: exponential error model t(y) = \log(y) and y = fe^{ae}
  • logit: logit error model t(y) = \log(\frac{y}{1-y})
  • band(0,10): extended logit error model t(y) = \log(\frac{y}{10-y})
  • band(0,100): extended logit error model t(y) = \log(\frac{y}{100-y})

Notice that in the two last cases, the representation as a function can be misleading but is easier to explain the band extension.

Autocorrelated observation error
Autocorrelation can be modeled for each observational model by selecting the autocorrelation option with the checkbox “r” in the Monolix user interface. When selecting the autocorrelation, Monolix estimates the autocorrelation parameter r. Note, that when several outputs are available then the autocorrelation can be selected for each output independently.

Positive gain on the error model
The second parameter b in the observational models comb1 and comb1c can be forced to be always positive by selecting b>0.

Mlxtran observational model syntax

The DEFINITION: block in the [LONGITUDINAL] section is used to define the observational model:

DEFINITION: 
observationName = {distribution = distributionType, prediction = predictionName, errorModel = errorModel(param)}

(notice that one can use type=continuous instead of distribution = distributionType)

For example, if the observation is a concentration with a combined error model (Concentration = Cc + (a+b*Cc)*e), the observational error model is written as

DEFINITION: 
Concentration= {distribution = normal, prediction = Cc, errorModel=combined1(a, b)}

When the observational error is defined in the Mlxtran model file, the user must declare the observational model parameters (a and b in the presented example) as inputs.

 

Rules and best practices

  • The eventual arguments of the error model can not be calculations, only input names.
  • In Monolix, the user can choose the error model through the interface.
  • In Monolix, the name of the error models input parameters can not have any name.
    • The name of the input should correspond to the definition of the error model (ex. a for a constant error model, b for a proportional error model, (a,b) for a combined1 error model, …)
    • If there are several continuous outputs, the names of the error models input parameters should be linked to the order of the outputs (1 for the first error model, …)
    • For example, for a single output, a combined error model writes without any number as follows
      DEFINITION: 
      Concentration= {distribution = normal, prediction = Cc, errorModel=combined1(a, b)}
    • For example, for two outputs, a combined error model and a constant error model write as follows
      DEFINITION: 
      Concentration= {distribution = normal, prediction = Cc, errorModel=combined1(a1, b1)}
      PCA= {distribution = normal, prediction = E, errorModel=constant(a2)}
  • Notice that a parameter can not be shared by two error model. For example, in the previous Concentration/PCA example, we can not replace a2 by a1.