[R-sig-ME] An advice needed on paper statistical analysis description

Ofir Levy levyofi at gmail.com
Wed Jul 29 13:42:50 CEST 2009


I will rephrase my question...
Does it reasonable to perform the statistical analysis using MCMC 
simulations instead of available well-tested and documented functions 
when p-values may mislead? If looking at the 0.05 order p-values. I had 
2 tests of 5 with p-values between 0.1 and 0.01 so I decided to perform 
the whole analysis with MCMC simulations. Under what kind of models I 
should also submit the BUGS code and detailed description of the models?
I know some of you have wonders sometimes what to write in your 
statistical description part of a paper so this might be a helpful 
discussion...

Cheers,
Ofir.

>

Highland Statistics Ltd. wrote:

>
>> I in the final stages of writing a paper which summarized some of my PhD 
> aspects. I did a lot of mixed effects modelling. I have started with 
> the lme function (I prefer it on `lmer` becuase it is more documented 
> and more flexible in terms of weight and correlation structures). I 
> came out with non-convincing p-values (As Zuur's 2009 book describes 
> it - between 0.001 to 0.1)
>
>


> Did we say that? We may have said that p-values in the order of 0.01 
> are not convincing for a GAM(M).
>
>
> Now I wonder what to write on the paper.
>
>
> We all do!
> Alain
>
>
>
> I guess I should explain why I used the somehow more complex approach 
> (writing the models myself in JAGS) and I hope the reviewers will not 
> reject the paper... I want to submit the paper to Ecology and most 
> papers in this journal with BUGS analysis are much more complex and 
> unique to a specific problem.
> I will be thankful to hear your advices.
>
>
>
>
> Thanks,
>
> Ofir.
>




More information about the R-sig-mixed-models mailing list