The RATS Software Forum

Dear Tom,
I am estimating a Panel using the pregress instruction with fixed time and country effects.
pregress(Method=fixed, effects=both, robusterrors, smpl=dummytest) y 1//1960:1 28//2015:01 resids
# ch_coreinf caput stur ch_hwcdw slack_econ ch_GDPq{1}

I would like to get the fitted values for each country and year . First I calculated the fitted values using the orignial series y and the value in the resids vector.
y+resids= y_fitted. There the results are resonable.
Then I wanted to calculate the fitted values by hand determining where the recent change in the fitted value from one year to the other comes from.
I used

prj fitted

When comparing the resids and the fitted values from the prj instruction, the fitted values seam strange.
.

print / fitted y resids

Is the prj instruction not accounting for fixed time and country effects?
How can I calculate the fitted values manualy to determine which variable contributed by how much to the change of the fitted value?

Thanks a lot in advance

PRJ only takes into account the "slope" coefficients---the fixed effects are considered nuisances. However, you can use the INDIV and TIME options on PREGRESS to retrieve those and then you can add them back one at a time or together.