This is not likely to happen, but if a value of an unused parameter is NaN, then factor(NaN*NA) here does not produce the right input for the map argument of MakeADFun.
I copied and edited validation.R and tmbprofile.R to deduce where the the map list element for the unused parameter was changed.
Reproducible Steps:
1) Make input initial parameter list with a value = log(-1) for an unused parameter.
2) Fit model
3) run TMB::oneStepPredict(method="oneStepGeneric").
Current Output:
Error in if (xnext + that < parm.range[1]) { :
missing value where TRUE/FALSE needed
(repeated for each observation)
Expected Output:
Normal execution of TMB:oneStepPredict: return data.frame with calculated residuals.
Possible Solution:
change referenced line to
map <- lapply(args$parameters[fix], function(x)factor(rep(NA,length(x))))
Description:
This is not likely to happen, but if a value of an unused parameter is NaN, then factor(NaN*NA) here does not produce the right input for the map argument of MakeADFun.
I copied and edited validation.R and tmbprofile.R to deduce where the the map list element for the unused parameter was changed.
Reproducible Steps:
1) Make input initial parameter list with a value = log(-1) for an unused parameter. 2) Fit model 3) run TMB::oneStepPredict(method="oneStepGeneric").
Current Output:
Error in if (xnext + that < parm.range[1]) { : missing value where TRUE/FALSE needed
(repeated for each observation)
Expected Output:
Normal execution of TMB:oneStepPredict: return data.frame with calculated residuals.
Possible Solution:
change referenced line to map <- lapply(args$parameters[fix], function(x)factor(rep(NA,length(x))))
TMB Version:
1.9.1, commit = e499559
R Version:
R version 4.1.3 (2022-03-10)
Operating System:
Windows 10