re: st: RE: RE: RE: unpaired regression

[David Airey <david.airey@vanderbilt.edu>]

> Assay1 is a high-throughput, simplified model for predicting the 
> response of
> the batch to assay2.  The replicates are randomized from the batch, so
> there's little systematic structure to the order of the replicates 
> (which is
> why I'm hesitant to pair them up).
I remain a little confused about the two different scales of the data. 
This suggests multiple response variables.

But batch is probably best treated as a random factor.

Collapsing the data as was suggested is probably not a great idea, 
because it might lead to confusion as to whether you measured the same 
batch replicate with two measures. My understanding was 6 separate 
replicates from each batch were randomly assigned between assay 
methods. If the scale of assay were the same this might be a two factor 
ANOVA with batch random and assay fixed, and replicate nested in batch# 
* assay# cell (the error?).

Anyway, since batch is a random factor, maybe what you want to do is to 
compare the intraclass correlations for the two methods? I wonder if 
you needed to measure the same batches with both methods to answer the 
question you wanted to ask, which seems to be variation within batch 
versus between batch, where you are hoping that the confidence interval 
for the IC of Assay1 is not different than Assay2? I dunno.

Let me know how it goes,

-Dave

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/