James W. MacDonald | 1 Jun 15:49 2009
Picon
Picon

Re: Random effects and variance components

Hi Paolo,

I don't think you can fit the model you describe using limma, and I 
really don't think you can get the variance components. If you want to 
fit more sophisticated mixed models, you will likely need to use the 
lme4 package, and the lmer() function in particular.

But note that this route will require much more work on your part, both 
in understanding how lme4 and lmer() work, as well as writing the code 
to fit individual models to each reporter and extracting the results.

Best,

Jim

Paolo Innocenti wrote:
> Dear Gordon and list,
> 
> thanks for the previous help, it was indeed helpful. Nonetheless, even 
> after some strolling (and independent trials-and-errors), I am still 
> stuck on this issue:
> 
> we came to the conclusion that the simplest good model for our affy 
> experiment is the following:
> 
> design <- model.matrix(~ sex*line, data=pData(data))
> 
> sex: M/F,
> line: 15 levels (different clones)
> 8 biological replicates for each line (4 for each sex)
(Continue reading)

Naomi Altman | 1 Jun 18:21 2009
Picon

Re: Random effects and variance components

Have a look at MAANOVA. --Naomi

At 09:49 AM 6/1/2009, James W. MacDonald wrote:
>Hi Paolo,
>
>I don't think you can fit the model you describe using limma, and I 
>really don't think you can get the variance components. If you want 
>to fit more sophisticated mixed models, you will likely need to use 
>the lme4 package, and the lmer() function in particular.
>
>But note that this route will require much more work on your part, 
>both in understanding how lme4 and lmer() work, as well as writing 
>the code to fit individual models to each reporter and extracting the results.
>
>Best,
>
>Jim
>
>
>
>Paolo Innocenti wrote:
>>Dear Gordon and list,
>>thanks for the previous help, it was indeed helpful. Nonetheless, 
>>even after some strolling (and independent trials-and-errors), I am 
>>still stuck on this issue:
>>we came to the conclusion that the simplest good model for our affy 
>>experiment is the following:
>>design <- model.matrix(~ sex*line, data=pData(data))
>>sex: M/F,
>>line: 15 levels (different clones)
(Continue reading)


Gmane