Question

Design matrix for missing group

0

Entering edit mode

ijvechetti ▴ 10

@ijvechetti-20701

Last seen 2.0 years ago

United States

Hello all,

I have an array with 8 groups and 4-time points (3 replicates in each), with the exception of 1 timepoint where the rats died. Reading about this missing group on Limma guideline, I think the best strategy would be the group-mean parametrization strategy?

Imaging I have groups (1,2,3,4,5,6,7 and 8) and time points (1mo, 3mo, 6mo, 12mo), how could I determine the changes over time doing the parametrization strategy, assuming I'm correct on that one.

Thanks in advance

limma • 1.0k views

ADD COMMENT • link written 4.9 years ago by ijvechetti ▴ 10

score 1 · Answer 1 · 2020-01-29

1

Entering edit mode

Gordon Smyth 52k

@gordon-smyth

Last seen 3 hours ago

WEHI, Melbourne, Australia

I generally recommend the group-means design matrix setup because of its simplicity and ease of interpretation. Whether there is a missing group or not makes no difference -- the approach is the same.

ADD COMMENT • link 4.9 years ago Gordon Smyth 52k

0

Entering edit mode

Thanks Gordon, for your quick answer. Coming back to my design, how could I determine changes over time doing this strategy?

Sample 1 is my control.

I did the contrasts for each month: Cont <- makeContrasts("2-1", "3-1", .......... levels=design)

Now I would like to determine if there is any change over time, but I'm not sure how to make the appropriate design. Would be something like that?

Cont.group2 <- makeContrasts((2-1)-(3-1)-(6-1)-(12-1), levels=design)

Thanks once again

ADD REPLY • link 4.9 years ago ijvechetti ▴ 10