R help - Oct 2010 - lme with log-normal distribution of parameters

Dear R-users,

Do you know if we can use the function lme in R for log-normal 
distribution of parameters as used in Nonmem ?
theta=theta0*exp(eta)

In our model, the parameters follow the log-normal distribution so it's 
not reasonable to deal with normal distribution which gives us negative 
values in simulation

Thanks for your help,

Thu



	[[alternative HTML version deleted]]

Hoai Thu Thai <hoai-thu.thai <at> inserm.fr> writes:
> 
> Dear R-users,
> 
> Do you know if we can use the function lme in R for log-normal 
> distribution of parameters as used in Nonmem ?
> theta=theta0*exp(eta)
> 
> In our model, the parameters follow the log-normal distribution so it's
> not reasonable to deal with normal distribution which gives us negative 
> values in simulation
> 
  No.
  There are various tools (AD Model Builder, WinBUGS, JAGS) which have
interfaces to R which could be used, but neither of the main R packages
(nlme or lme4) will handle non-normal random effect distributions
(it's possible that some of Jim Lindsey's packages at 
<popgen.unimaas.nl/~jlindsey/rcode.html> will handle this case;
I don't remember).

   You might want to try this question on r-sig-mixed-models at r-project.org

  Ben Bolker