simulateResidual error - incorrect number of dimensions

[NLJames]

Hi @florianhartig,

Firstly, DHARMa is amazing, thank you for all the hard work.

I am having an issue I have never encountered after hundreds of uses of this function. I ran a model in glmmTMB which I have checked has converged with a sensible output. But running simulateResiduals() on this object does not seem to be able to interpret the output. Strangley, I have never had this issues before with glmmTMB model objects. Can you shed any light on why this is happening? I can share the data if need be.

m4<-glmmTMB(resids ~ wSpeed_mn + (1|BandNumber)
            , family = gaussian()
            , data=noNA_resiDat
            , REML=TRUE)

pRes1<-simulateResiduals(fittedModel =  m4, n=2000)

Error message

Error in mr[, 1] : incorrect number of dimensions

[NLJames]

I have figured out the issue here, posting info for posterity. Seems that the resids column was stored as a matrix within the data frame used for the model. It was not converted through the glmmTMB function which I suppose was causing problems when handed to DHARMa. So seems that the column classes are important even at the modelling stage.

[melina-leite]

Hi @NLJames, thanks for reporting and finding the error!
This is related to issue #280, but now with glmmTMB.
This is a bug that should be fixed soon.

Cheers

[florianhartig]

Hi @NLJames,

as Melina said, we'll be looking at this - for this, and just if it's no inconvenience, it could be useful if you could provide a reproducible example of your problem and post it here, just so that we can check if a possible solution to the issue fixes the problem?

Cheers
Florian

[bbolker]

If this is something that triggers any problems in glmmTMB (as opposed to being DHARMa-specific) please post an issue for glmmTMB as well.

[bbolker]

A reproducible example of a warning (not an error):

mtcars$mpg <- scale(mtcars$mpg)
m <- glmmTMB::glmmTMB(mpg ~ cyl, data = mtcars)
DHARMa::simulateResiduals(m)
Object of Class DHARMa with simulated residuals based on 250 simulations with refit = FALSE . See ?DHARMa::simulateResiduals for help. 
 
Scaled residual values: 0.548 0.564 0.116 0.612 0.912 0.208 0.44 0.272 0.112 0.292 0.16 0.68 0.772 0.548 0.076 0.076 0.504 0.972 0.9 1 ...

Warning messages:
1: In securityAssertion("nKcase - wrong dimensions of response") :
Message from DHARMa: During the execution of a DHARMa function, some unexpected conditions occurred. Even if you didn't get an error, your results may not be reliable. Please check with the help if you use the functions as intended. If you think that the error is not on your side, I would be grateful if you could report the problem at https://github.com/florianhartig/DHARMa/issues

Context: nKcase - wrong dimensions of response
2: In securityAssertion("nKcase - wrong family") :
Message from DHARMa: During the execution of a DHARMa function, some unexpected conditions occurred. Even if you didn't get an error, your results may not be reliable. Please check with the help if you use the functions as intended. If you think that the error is not on your side, I would be grateful if you could report the problem at https://github.com/florianhartig/DHARMa/issues

Context: nKcase - wrong family

(Adding a random effect and REML=TRUE doesn't seem to make a difference.)

[melina-leite]

Hi @bbolker, thanks for sharing an example with the warnings. It was an issue with the function getObservedResponse.default() that was expecting a matrix for the observed responses only for n/k binomial models. We no longer have a warning about "binomial" problems when the response variable is a matrix (as when using scale()).
So, I closed issue #280.

However, I'm not sure anymore if this issue problem is related to it. @NLJames, if possible, could you please provide a reproducible example of your case? It would be also great if you could try to run the code with the latest DHARMa development version to check if the solution to issue #280 also worked for your problem. :)

Best,
Melina