Tianchen Xu
Tianchen Xu
`nfc-mfclassic W a $TMPFILE_MFD` $TMPFILE_MFD is the dump file.
That's strange. I am using the github version. I just did the following things several minutes ago and got the same result. ``` > remove.packages('KFAS') # remove any previous KFAS...
I just tried R 4.0.2 with clean install and get the same result. ``` > logLik(model) [1] 101060.8 > version _ platform x86_64-w64-mingw32 arch x86_64 os mingw32 system x86_64, mingw32...
Yes. I checked many times that I am using the same image in dropbox. I just run your codes in the lastest reply: ``` > diag(model$P1inf)[1:4] diag(model$P1)[1:4] logLik(model) # not...
They are all on Win 10 x64. This is what I get: ``` > logLik(model) [1] 101060.8 > logLik(model, H_tol = 1e300) [1] 101060.8 > logLik(model, H_tol = 1e500) [1]...
I didn't modify Makevars. When I run ` install_github('helske/KFAS')`, this is the complete log: ``` > install_github('helske/KFAS') Downloading GitHub repo helske/KFAS@master √ checking for file 'C:\Users\Peter Xu\AppData\Local\Temp\RtmpuuNMQz\remotes3318641250dc\helske-KFAS-e04e6b7/DESCRIPTION' ... - preparing...
Btw I find something. After I load the R image: ``` > logLik(model) # wrong [1] 101060.8 > logLik(approxSSM(model)) # this is reasonable [1] -5357.131 > approxSSM(model)$iter [1] 13 >...
Wow, that's great! You can reproduce the issue. Thank you!!
> It is likely that your model is just poorly identifiable which results numerical issues during the filtering/smoothing. Yes, it is possible that the model is poor with these parameters,...