Entering edit mode
Hinrich Göhlmann
▴
20
@hinrich-gohlmann-845
Last seen 10.2 years ago
Good morning,
I have a suggestion for affycomp and it's accompanying web site at
http://affycomp.biostat.jhsph.edu/AFFY2/TABLES.hgu/0.html
I very much like the idea to check how the different algorithms
perform
but I would even more appreciate a biologist-compatible interpretation
of the comparison table (low or no stats!). What I have in mind is
something like:
(1) define all the different approaches that people could be
interested
in from a statistical point of view (Example: I am interested in a
best
possible accuracy of fold change estimation at all signal levels. For
this application, score components 4, 7, and 13 are most important (I
don't know!). OR: I would like to have the highest possible precision
for the signal levels. For this application score components 3, 6, 8,
and 11 are most important. OR: I need best possible precision for low
expressed genes. For this components x, y, and z are most important.
And
so on...
(2) make a table with those applications and give the 2 or three best
suitable algorithms for those applications.
By openly defining those applications at such a web site, we can get a
discussion going on what components most people agree on per
application
and can secondly make practical use of the data that gets collected.
Cheers,
hinrich d8-)
