HPC and stack_modelling

Hi there,

I have been running stack_modelling on my Mac in RStudio without problem but now I am trying to send off a larger model through SSH to an HPC (GCC 8, R 4.0.0-cairo).

My model looks like:

Plant_ssdm <- stack_modelling(c("MARS","CTA"), Plant_occ, Env, rep = 10, 
                              Xcol = "longitude", Ycol = "latitude", Spcol = "species",
                              ensemble.thresh = 0.5, ensemble.metric="AUC", cv="LOO",endemism="WEI",
                              uncertainty = TRUE, method = "PRR.pSSDM", cores=0, verbose=TRUE)

Ensemble_modelling is working but when I try to run stack_modelling I get:

Error in Occurrences[which(names(Occurrences) == Spcol)] == species : 
  comparison of these types is not implemented
In addition: Warning message:
In eval(e, x, parent.frame()) :
  Incompatible methods ("Ops.data.frame", "Ops.factor") for "=="

I got Plant_occ through load_occ and the file itself looks alright.:

> head(Plant_occ)
    X latitude longitude                 species
11 11  -9.2500  147.3333 Syzygium branderhorstii
13 13  -6.4167  145.7833          Melicope rubra
14 14  -6.0000  144.7500   Eragrostis tenuifolia
16 16  -6.3333  147.1667      Myosotis australis
19 19  -5.1167  141.6333        Drimys membranea
20 20  -3.5833  142.7667      Polygala glomerata

Do you happen to have any idea what might be happening?

Many thanks and happy holidays,
Juniper

Hi Juniper,

I can only guess that there are some data type problems in the part where the data is subset for each species. I tried running the code in question (lapply over species)

SSDM/R/stack_modelling.R

Line 435 in d8152bd

    
           Spoccurrences <- subset(Occurrences, Occurrences[which(names(Occurrences) == Spcol)] == x)

but could not reproduce the problem.

You said stack_modelling worked on your computer? Are you running the same R version on the cluster? I have not updated to R 4.0.0. yet and I can imagine there have been changes that are in conflict with the data.frame == factor comparison we did.

Hi Juniper,

I can only guess that there are some data type problems in the part where the data is subset for each species. I tried running the code in question (lapply over species)

SSDM/R/stack_modelling.R

Line 435 in d8152bd

Spoccurrences <- subset(Occurrences, Occurrences[which(names(Occurrences) == Spcol)] == x)

but could not reproduce the problem.
You said stack_modelling worked on your computer? Are you running the same R version on the cluster? I have not updated to R 4.0.0. yet and I can imagine there have been changes that are in conflict with the data.frame == factor comparison we did.

Hi Lukas,

Many apologies - I only saw your reply now.
I had to change R versions and installed SSDM directly from Github and then it worked!
Many thanks again!