I recently brought up this issue in the programming thread, but now that I have a clearer view of how eviews treats this topic. I wanted to share it in this forum, so if anyone is interested, please drop a line here.
It's impossible to standardize coefficients if the model has AR/MA terms, but I have tasted the forbidden fruit by manually standardizing every data series, then added my desired AR/MA terms. For referenced I standardized by:
Code: Select all
series x1_standard = (x1 - @mean(x1))/@stdevs(x1)
So now I have standardized coefficients if I wanted to make direct comparisons between them, and I have my AR/MA terms. I am just wondering now how much I want to invest in trying to interpret the AR/MA terms, and if it's fair to directly compare them with the non-AR/MA terms, even in this unitless paradise I have created.
I also noticed they have taken on different roots and impulse responses when compared to my raw data.
So if any one else has ever fiddled around with this kind of thing, feel free to comment/inquire.
