meanVarPlot Ignores Sample Total Count Differences
0
0
Entering edit mode
Dario Strbenac ★ 1.5k
@dario-strbenac-5916
Last seen 2 days ago
Australia

I think the example data set below should produce a variance of 0 across the range of means but it doesn't because the library sizes aren't adjusted to be comparable between samples. An adjustment also doesn't happen for the yeast data set on page 8 of the vignette.

counts <- matrix(c(seq(100, 1000, 100),
seq(100, 1000, 100) * 2,
seq(100, 1000, 100) * 3), ncol = 3)

meanVarPlot(newSeqExpressionSet(counts)) # Appears as over-dispersion.

Should the meanVarPlot function require that normalizedCounts has been provided to the newSeqExpressionSet constructor and the EDASeq vignette be updated to demonstrate a library size correction?

EDASeq meanVarPlot • 360 views
0
Entering edit mode
I don't think it's reasonable to *require* normalized counts, as one might want to look at the mean variance relation of raw counts. If the object contains normalized counts, meanVarPlot will use them.