Design Matrix for Limma

Question

[hakimelakhrass]

So I am trying to make a design matrix for limma. I need to make sure it is a paired analysis because it is time series data. How ever I am not sure I am doing it right.

filenames <- row.names(pData(rawData))
person <- factor(c((seq(1,13, by=1)),seq(1,14,by=1),seq(1,14,by=1)
                   ,seq(1,14,by=1),Filter(function(x) x != 8 & x != 9, 1:14)))

time <- factor(c( rep.int(1, 13)),rep.int(2,14),rep.int(3,14)
                 ,rep.int(4,14),rep.int(5,12)))

design <- model.matrix(~0 + person + time )
rownames(design)<- filenames

This is the matrix I make. I input into the model and my probelist is super long and does not look right. I want to compare between the 5 times points but at the same time take into account the that they are the same person. Any help?

Thanks.

[andrew.j.skelton73]

Try reading the Limma users guide section 9.4.1 for paired model designs, it'll give you some worked examples to look at.

[hakimelakhrass]

I did do that. My example is based on that section. They only had one example though.

[russhh]

Could you edit your definition of time so that it compiles, please:

time <- factor( c( rep.int(1,13), rep.int(2,14), rep.int(3,14),
              rep.int(4,14), rep.int(5,12)))

I was wondering why you are using time as a factor. Rather than trying to perform contrasts at each time point, I'd try and find time-related trends using polynomial regression.

[Devon Ryan]

This appears to be a rendering bug in biostars. Everything renders correctly in the preview but not the final page. I'll play around with things until they render in the final version in a valid manner.

[hakimelakhrass]

That is just how I saw it being done. To make a contrast matrix and try to see the differences. Do you have any tutorials or R packages in using polynomial regression for this?