Can we get ga() compatibility? ga() from gamlss.add package

[pntoiv]

Hey!

I am using ggeffects 1.7.2.

I tried to create predictions with ggpredict but I ran to this error:

> P <- ggpredict(ModGAMLSS, terms=c("VAR1"), back_transform=T)

Error in names(x) <- value :
'names' attribute [10] must be the same length as the vector [4]

I am using gamlss-package and ga() function to build my model, and my model is something like this:
gamlss(Response~X+ga(s(VAR1)+s(VAR2)+te(VAR3, VAR4)), sigma.fo=~VAR4, data=data, family=LOGNO2)

I believe this is because I used ga() function from gamlss.add package to fit my smooth terms. https://www.rdocumentation.org/packages/gamlss.add/versions/5.1-13/topics/ga

When I use predict(model, type="terms") it is not able to separate the terms in ga() into their own columns. I can extract gam model object with getSmo(model) and when I do that, it is able to separate the smooth terms into their own columns, but ggpredict() doesn't work with the object returned by getSmo():

> P <- ggpredict(getSmo(ModGAMLSS), terms=c("VAR1"), back_transform=T)

Error in model_data[, rn, drop = FALSE]:
! Can't subset columns that don't exist.
✖ Column Y.var doesn't exist.

I really like using ga() when I want to bring my gam() model from mgcv-package to gamlss package.

I wonder if there is a possibility to get ga() compatibility with ggeffects?

[strengejacke]

Do you have a reproducible example?

[pntoiv]

Do you have a reproducible example?

Well, confusingly I edited and subset my data to make a dataset for reproducible example and then it worked (though couldn't compute confidence intervals), so I have no idea why my original data doesn't work. So it must has something to do with the number of variables in the model, as I reduced the number of variables for the subset data.