One comparison that I did was to run the model (v1.10 pre-release version) where the missing data in the expression sheets was filled in with the average and where the data were left as missing.
I found big discrepancies in the parameter values estimated, not the zero or minor differences that I was expecting.
I will write up more specifics of the results and come up with ways to test what is going on.
I have been working on reproducing results of db1-7 input workbooks (data are found in the Dahlquist Lab repository here: https://github.com/kdahlquist/DahlquistLab/tree/master/data/Fall2018).
One comparison that I did was to run the model (v1.10 pre-release version) where the missing data in the expression sheets was filled in with the average and where the data were left as missing.
I found big discrepancies in the parameter values estimated, not the zero or minor differences that I was expecting.
I will write up more specifics of the results and come up with ways to test what is going on.