Study cases

[Lluís Revilla]

I was doing the cases as in the website, when I noticed that they aren't run with the latest version of mixOmics.

The case I have trouble with is the sPLS

network(liver.spls, comp = 1:2, shape.node = c("rectangle", "rectangle"),
        color.node = c("white", "pink"), color.edge = color.edge, threshold = 0.8)
Error in network(liver.spls, comp = 1:2, shape.node = c("rectangle", "rectangle"),  : 
  unused argument (threshold = 0.8)

It only works when I remove the threshold parameter:

network(liver.spls, comp = 1:2, shape.node = c("rectangle", "rectangle"),
        color.node = c("white", "pink"), color.edge = color.edge )

Maybe it would work better to run these cases as vignettes inside the package. That way the code and content will be up to date.

[Lluís Revilla]

There is another tutorial that I'm having problems, it reports some warnings which make me doubt of the results

perf.srbct <- perf(splsda.srbct, validation = "Mfold", folds = 5,
                    dist = 'max.dist', nrepeat = 10,
                    progressBar = FALSE) 
Warning messages:
1: The SGCCA algorithm did not converge 
2: The SGCCA algorithm did not converge 

[Kim-Anh Le Cao]

Thanks Lluis we will fix that on the website. Many things to do on our end and we running at a stretch.

There might be some convergence issue using perf, it depends on the random folds. I would not be worried though as this happened (here) twice in a 5 folds x 10 repeats = 50 runs and we average the results. I would be concerned however if that happened on a splsda() final model though (which is not the case here).

I am closing this issue.

[GENET Carine]

The argument Threshold was replaced by cutoff. so the spls command should be :

network(liver.spls, comp = 1:2, shape.node = c("rectangle", "rectangle"),
        color.node = c("white", "pink"), color.edge = color.edge, cutoff= 0.8 )

Regards Carine Ca