Hi,
I wonder after running fitPLM, besides coeff and standard errors, what
else
are computed in the procedure? How can I check it?
For the image, how can I turn on the legend? The documentation
mentions
turn on the add.legend which I'm not sure how to do it.
And, the command in line -3 pg3 of AffyPLM is cut off, can someone
give me
the complete command?
Thanks.
Yen Lin
> For the image, how can I turn on the legend? The documentation
mentions
> turn on the add.legend which I'm not sure how to do it.
Perhaps something like
image(Pset,add.legend=TRUE)
> And, the command in line -3 pg3 of AffyPLM is cut off, can someone
give me
> the complete command?
Pset <- fitPLM(Dilution,model=PM~-1+probes+logliver+scanner,
normalize=FALSE,
background=FALSE,
variable.type=c(logliver="covariate"))
Lawrence,
In my experience of using the images its very easy to see artefacts
such
as dirt or air bubbles as these show up as green on the images.
Recently
I also detected a green strip across a certain batch of chips and I'm
currently looking into whether there could be a scanner problem. Also
a
'bad' chip could appear darker overall. Ideally I guess you're looking
for
a random distribution of colours on a chip and a similar overall
colour
between chips. In general chips that look quite bad still seem to give
resonable data, I don't know what the point is when you exclude a
chip.
As for the residuals vs. SE perhaps this poster might help.
http://stat-
www.berkeley.edu/users/terry/Group/talks/Aug2003/QCPoster.pdf
As a secondary point is the NUSE discussed in the poster the same as
the
model SE that you can boxplot in affyPLM, as the boxplots are adjusted
so
each probeset has a median of 1?
I'd also appreciate a lay-mans explanation of what the output of PLM
and
what it can be used for. For example can the SE's be used in any way
to
get some kind of confidence for differential expression?
Cheers,
Matt
