Issues with plot.perf_mddsPLS

[HHoskens]

Hello,

I am trying to do CV to select lambda for my PLS model, but I get an error when trying to run perf_mddsPLS (can be traced back to the plot.perf_mddsPLS function):

Error in xtfrm.data.frame(x) : cannot xtfrm data frames

The mddsLS function works fine with the same X and Y input variables. I am working in R 4.3.1.

Thanks!

[hlorenzo]

Hello,
Thx to try ddsPLS, what is the nature of your matrix ? All numeric ?

[HHoskens]

Correct. Both are numeric.

[hlorenzo]

Can you send me a copy of you df, in private and after noising data ?

[HHoskens]

It doesn't seem to be data specific; I get the same error when running the "liverToxicity" example. Might be related to the "order" function and some R versions

[hlorenzo]

Actually the function you try to use is not maintained anymore. I am sorry you had to deal with it.
Consider using the ddsPLS function itself such as

library(ddsPLS)
data(liverToxicity)
X <- scale(liverToxicity$gene)
Y <- scale(liverToxicity$clinic)
model <- ddsPLS(X,Y,doBoot = TRUE,n_B=10,verbose=TRUE)

You will then perform bootstrap operations in order to get variability. In the console you will get something like:

and on the plot area:

Here the criterion minimized is the difference between R2-Q2. You can consider minimizing Q2 if you wish through the argument criterion. You can also consider the plot function if you wish to explore the built weights (through the argument type put equal to "weightX" or "weightY".

Thank you for reminding me that I must consider rewriting this vignette, I will tell you as soon as this is done. If you have any other question, do not hesitate to ask, I hope you will find interesting stuff in your data

[HHoskens]

Ok, I will give this a try. Thank you for your help!

and just to be sure, ddsPLS should be downloaded from CRAN (and not github)?

[hlorenzo]

On GitHub, favor ddsPLS2 rather than ddsPLS. I work on the version to the CRAN. I will let you know when I have done my job

[hlorenzo]

The new version on the CRAN should fix your bug. Please test and tell if everything is ok