2 replies to this topic

[Christophe Boetsch]

Hi,

 

While running a popPK model with NLME, I do sometimes get the variability being calculated for the theta, sometimes not.

 

Any idea as to why this happens?

 

Using PHX 1.2, NLME 1.1

[Samer Mouksassi]

Variability can mean different things:

residual error

between subject variability

uncertainty ( standard error on estimate)

 

I assume you meant the standard error of estimates.

Most of the time it is because the hessian cannot be inverted ( overparameterization , local minimum, ) try to fix a parameter or change your initial estimate.

 

Samer

[Simon Davis]

Christoph, as Samer indicated, I sometimes find that 'accepting all' on the final estimates and then re-executing will give me the SE.

 

  Another thing to consider is the algorithm you have selected; the FOCE-ELS often gives better results for many problem types that the -LB variant.

 

 Simon