Matrix column bound exceeded when using generalized size comp

[N-DucharmeBarth-NOAA]

Has anyone run across this particular error before? I can't seem to find where in the source code that this error would be triggered.

...
Evaluating biology calculations once ... done
Starting optimization of 'ss3_win' in phase 1 of 1 at Wed Jan 14 13:46:55 2026
matrix column bound exceeded -- column index 1 is higher than 106 in linad99\fvma_acc.cpp:49.

I am playing around with a toy model (single area, 14 fleets, 1 index, length comp & weight comp), and best as I can tell there seems to be an interaction between the generalized size comp (weight data) and the population bin structure. There are 106 pop length bins, and I only get this error when the generalized size comp data is used. The model runs fine when only the index and/or the length comp is included but produces the error if I turn on the size comp data. I have found a work around where I increase the pop length bins and I truncate the number of data bins in the size comp data but I'm not sure why that works and it is unsatisfying to have an inefficient model that throws out some of my data. This seems to be a different error than the one that you get when the size comp bins are narrower than the population length bins.

[Rick-Methot-NOAA]

Hi Nick,
This is an ADMB error code, so very hard to diagnose which matrix is being violated. I suggest starting by looking at the warning file and then echoinput to see if anything stands out. If not, then pack up the input files and send to nmfs.stock.synthesis@noaa.gov so we can run them and look at the details.

[N-DucharmeBarth-NOAA]

Hi @Rick-Methot-NOAA, thanks for getting back to me. I had a closer look at the model and it definitely seems to be something with the generalized size comp (weight data). When I turn off that data component and just fit to CPUE and length comp the model appears to run just fine. From the echo input it looks to "crash" the first time it reads in any of the weight data. It doesn't seem to matter which year/fleet of weight data it is as it still crashes on the first read even if I remove the first several observations. I'll go ahead and share the files. Like I said before, I can get it to run if I go from 2cm population length bins to 1cm but I'd rather use 2cm bins since the model is way faster with the reduced dimensionality of the population.