"param" versus "coefficient.fill"

[fioramanti]

I would initialize a SSpace giving the model some starting values. The natural way would be by using the “param” command. The problem with this is that (my understanding is) it can only take numbers.
My starting values are the results of some calculations, so I prefer (to have all the stuff automatic) to give names (“scalar” to be precise) to this values. In this way I can’t use “param”.
My take is that I can fill (“mycoefficientsvector.fill”) the coefficient vector eViews creates once a new wf is created. And this is exactly what I did and it works.
The point is that the final coefficient estimates are different (even if the differences are not so large) if I use the first or the second option.
Can anyone explain me why? What is the difference between the two methods?
Thanks,
Marco

[EViews Glenn]

There shouldn't be differences. Can you provide a workfile with an example?

[fioramanti]

Yes, Attached.


Marco

[EViews Glenn]

Where is the param example? When you used param, to what precision did you enter the values?

[fioramanti]

The line with param (commented) is just above coefficient.fill. uncomment the first and comment the second. It is just one prg with both options. I know it could be (in part) a matter of approximation, but it shouldn' produce this difference (in think). The precision is as much as I can read from the reference paper.