[R-sig-ME] Hi listers

[Malcolm Fairbrother]

Hi Said,

In the online supplementary information for my article available here (
https://doi.org/10.1017/psrm.2013.24), you can find R code for such
simulations, testing the impact of a skew-normal distribution for the
random intercepts.

Co-authors and I also tried a Chi-sq distribution in simulation studies for
this paper:
https://www.researchgate.net/publication/299604336_Fixed_and_Random_effects_making_an_informed_choice
I can send you R code for that too, if you're interested.

Others may be interested to know that non-normality doesn't seem to make
much difference.

Best wishes,
Malcolm


Dr Malcolm Fairbrother
Reader in Global Policy and Politics
School of Geographical Sciences  •  Cabot Institute  •  Centre for
Multilevel Modelling
University of Bristol





Date: Tue, 25 Apr 2017 13:07:14 +1000
> From: Andrew Robinson <mensurationist at gmail.com>
> To: "=?utf-8?Q?R-sig-mixed-models=40r-project.org?="
>         <R-sig-mixed-models at r-project.org>, Said Ali Shah
>         <said_lect06 at yahoo.com>
> Subject: Re: [R-sig-ME] Hi listers
>
> Generate them all from the Uniform distribution and then transform them to
> the distribution of your choice.
>
> Cheers,
>
> Andrew
>
> On 25 Apr 2017, 1:00 PM +1000, Said Ali Shah via R-sig-mixed-models <
> r-sig-mixed-models at r-project.org>, wrote:
> > i am conducting a simulation study regarding sample size using
> multilevel models for longitudinal design, and want to see the effect of
> non-normal level-II errors now the question is 1. whether i should generate
> the level-II errors from normal distribution after specifying parameters
> which are required for the model and then transform these normal errors to
> non-normal errors? if yes then how to transform them to exponential and
> log-normal distribution2. i should generate these errors separately from
> normal as well as from exponential and log normal distributions. keep in
> mind that both results i.e., for normal and non-normal errors are required.
> Thanks in advance Regards,
> >
> > Said Ali Shah
>

	[[alternative HTML version deleted]]