Creating quintiles and running regressions on each

[gerrard]

Hi,

I have a panel. I want to estimate a simultaneous equations model (2 equations) using 2SLS.
My DV is ROA (performance), my IV is LEV (leverage), and vice versa in the 2nd equation. Both equations contain other variables such that they are identified.

Now I want to divide the panel into quintiles and estimate the equation system for each of those quintiles. The quintile division shall be done based on LEV, in such a way that there is an "around"-median quintile and 2 below- and above-median-quintiles, respectively.
So: Assume the median for LEV is 0.45. The max is 0.8, the min is 0.1. Then the quintiles should be:
1: >=0.1, <=0.24
2: >0.24, <=0.38
3: >0.38, <= 0.52 (so a range 0.07 around the median)
4: >0.52, <=0.66
5: >0.66, <=0.8

My questions:
1. How do I divide the panel into quintiles based on the median value that I calculated for LEV in the way described above?
2. How do I then run my 2SLS regression for each of those quintiles separately?

Thanks!

[EViews Gareth]

Use the estimation sample box. Something like:

Code: Select all

@all if lev>=0.1 and lev<=0.24

and so on. You'll have to re-estimate the same system with different samples, or use 5 different system objects. I'd chose the later.