Hi, it is my first time in the forum.
I have a question about VIP (Variable Importance for Projection) in PLS-DA analysis. Do you know whether it is possible to do this graph in MixOmics?
VIP is shown by group and not by component or to overall model.
Thanks very much.
Hi @enzo and welcome!
Unfortunately, we only provide the VIP for each variable and up to a component, as given in example(vip) (with a barplot).
data(linnerud)
X <- linnerud$exercise
Y <- linnerud$physiological
linn.pls <- pls(X, Y)
linn.vip <- vip(linn.pls)
barplot(linn.vip,
beside = TRUE, col = c("lightblue", "mistyrose", "lightcyan"),
ylim = c(0, 1.7), legend = rownames(linn.vip),
main = "Variable Importance in the Projection", font.main = 4)
You might be able to tweak this output for your needs. At the moment we do not have more capacity for such enhancements.
Kim-Anh
Thanks very much for your response.
But, theoretically is it possible calculate a VIP by variable by group for whole model?
I don’t find paper with this plot type except this one.
Hi @enzo,
Per model: yes you would consider only the last component value.
Per group: I assume you mean per sample group. I dont see how this can be calculated, here are the formula we use, feel free to tweak the vip() function for your own needs.
From our current handbook (in preparation):
Kim-Anh
Ok,
me neither, I don’t see how this estimate per sample group can be calculated.
Another question (I’m sorry for asking so many questions). Could I calculate Q2, R2 with MixOmics?
Thanks again
Enzo
hi @enzo
We have the R2 and Q2 for a PLS object using the perf() function. We have not implemented it for PLS-DA. You could do it manually by inputing Y as a dummy matrix (Y.dummy <- map(Y)) then input in pls(X, Y.dummy), although we never really tested. Q2 and R2 seem best defined for a regression model rather than classification.
Kim-Anh
