[BioC] Re: Genechip experiment in limma

[Gordon Smyth]

At 03:17 AM 24/08/2003, Christian Probst wrote:
... deleted stuff
>A finally question, as I don´t know your opinion about.
>
>  I am not a statistician. I would like to know if my point "it is better 
> to have more time points per experiment than more replicates per time 
> point and less time points per experiments" is a good one.
>
>Because I am "replicating" my data, but collecting another kind of info 
>(temporal info) also.

Well, collecting more time points isn't replication. You're asking whether 
it is better to have more time points *instead* of having replication. As 
always, what's a good design depends on the what questions you want to 
answer and what analysis you're going to do.

If you have lots of time points so that changes between consecutive time 
points will be small, then more time points might be better than 
replication. If you're just going to compare expression levels without any 
formal statistical analysis, or you're going to use an exploratory or 
unsupervised data technique like cluster anlysis, then again more time 
points might be better. If however you want to use statistical inferential 
techniques to assess differential expression, and your time points are such 
that genes may substantially change their expression levels between 
consecutive times, then I cannot see any alternative to having true 
replication.

Regards
Gordon

>Anyway,
>
>Thanks for your patience. It is always a pleasure to read your comments in 
>the BioC list
>Christian