V gone -ve definite for block 0 iteration aborted

Welcome to the forum for R2MLwiN users. Feel free to post your question about R2MLwiN here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Go to R2MLwiN: Running MLwiN from within R >> http://www.bris.ac.uk/cmm/software/r2mlwin/
Post Reply
tjsduq64
Posts: 31
Joined: Mon Jul 15, 2019 10:04 pm

V gone -ve definite for block 0 iteration aborted

Post by tjsduq64 »

I have a two-level random slope model shown below. If I run it in R with runMLwiN, I get an error saying "V gone -ve definite for block 0 iteration aborted(fixed pass)" after about 10 minutes. If I use "debugmode=T" in the estoptions, I get MLwiN window without anything on it, and the software crashes after about 10 minutes and gives the same error.
One odd thing is that if I add more variables to the fixed part, the model runs fine.

Code: Select all

options <- list(debugmode = T, Meth = 0, resi.store = T, residoptions='standardised',
                weighting=list(weightvar=c(NA,"weight"), standardised=T), reset=c(2,2))
model <-  life_satisfaction ~ 1 + gdp_c +
  (1 + gdp_c| country) + (1| id)
nullRS_ls_gdp <- runMLwiN(Formula = model, data = sample, estoption = options) # complete case analysis
summary(nullRS_ls_gdp)
Anyways, to run the model with the only variable gdp_c in the fixed part, I take gdp_c out of random part and ran the model. This works. Then, I used "debugmode=T", run the model without the gdp_c in the random part, manually add gdp_c in the random part, and click "More". This works but when I get an error "wrong number of output columns".

In summary, I want to run the model above but if I run it debugmode=F, it doesn't work. If I run it debugmode=T, it doesn't work and doesn't give me a chance to manipulate the model. It only works when I run it debugmode=T with another model and then manipulate it but then I have wrong number of output columns.

Could you give me advice?

Very much related question is what is this error about crashing in about 10 minutes without showing anything on MLwiN window? It doesn't give me a chance to manipulate the model, and this has happened to me for some other models as well...

Sun

tjsduq64
Posts: 31
Joined: Mon Jul 15, 2019 10:04 pm

Re: V gone -ve definite for block 0 iteration aborted

Post by tjsduq64 »

By the way, I thought the idea of using "starval" to set starting values was similar to clicking "More" after manipulate the model so I tried it but it doesn't help.

ChrisCharlton
Posts: 1165
Joined: Mon Oct 19, 2009 10:34 am

Re: V gone -ve definite for block 0 iteration aborted

Post by ChrisCharlton »

Without the data it is hard to see what is causing this. Do you have any missing values? Is it possible that your gdp_c variable is constant within any of your level-2 units? The startval option is similar to setting starting values and using the "More" option, however it will still run the model for a couple of iterations prior to applying these to set up the model structure, so it's possible that the problem is occurring before these values are set.

tjsduq64
Posts: 31
Joined: Mon Jul 15, 2019 10:04 pm

Re: V gone -ve definite for block 0 iteration aborted

Post by tjsduq64 »

I don't have missing values for gdp_c, and also gdp_c is constant within my level 2 units. It is a level 2 variable so each level 2 unit has a single value for all level 1 units within. This is analogous to modeling complex level 1 variation (i.e., modeling level 1 variance as a function of level 1 variable) but applied to level 2 (modeling level 2 variance as a function of level 2 variable). Not sure why this is an issue for this model because for other models it works fine. Especially in this case it's odd because if I add more variables in the fixed part, the model works, or if I manipulate it in MLwiN, it works.

ChrisCharlton
Posts: 1165
Joined: Mon Oct 19, 2009 10:34 am

Re: V gone -ve definite for block 0 iteration aborted

Post by ChrisCharlton »

Does it make any difference if you remove the reset and/or weighting options?

tjsduq64
Posts: 31
Joined: Mon Jul 15, 2019 10:04 pm

Re: V gone -ve definite for block 0 iteration aborted

Post by tjsduq64 »

That did not fix the problem. I also tried the same model in other mulitply imputed dataset to see if it only works in some datasets, but it did not work in any.

ChrisCharlton
Posts: 1165
Joined: Mon Oct 19, 2009 10:34 am

Re: V gone -ve definite for block 0 iteration aborted

Post by ChrisCharlton »

This does feel like it is having trouble getting started. If you were able to install from source I would suggest changing the number of initial iterations specified by the MAXI command on line 1384 of https://github.com/rforge/r2mlwin/blob/ ... ite.IGLS.R to either one or zero to see whether that makes it behave any better.

Post Reply