Re: [Gretl-users] loop
On Wed, 7 Sep 2016, Gmail wrote:
I've got a big issue to solve. I must implement an event study on
a sample
of 100 firms.
In order to avoid to implement 100 regressions I try to use a loop:
"list ylist = 1 3 5 7 9 \
11 13 15 17 19 \
[...]
191 193 195
list xlist = 2 4 6 8 \
10 12 14 16 18 \
20 22 24 26 28 \
[...]
190 192 194 196
loop foreach i ylist --progressive
loop foreach j xlist --progressive
ols ylist.$i const xlist.$j
genr alpha = $coeff(0)
genr beta = $coeff(xlist.$j)
store vectors.gdt alpha beta
endloop
endloop
open vectors.gdt
ylist and xlist are the lists that contain my independent variables.
Unfortunately, such code regress the first dependent variable on the other
100, then the second on the same 100 and so on and so forth.
How I can write a loop that only regress the variable 1 on the variable 2,
the variable 3 on the 4 etc?
How I can store the coefficients?
Something like this:
<hansl>
scalar N = 100 # adjust as needed
matrix B = zeros(N, 2)
loop k=1..N --quiet
j = 2*k
i = j-1
list L = i 0 j
ols L --quiet
B[k,] = $coeff'
endloop
print B
</hansl>
The matrix B will hold the coeff vectors as rows.
Allin Cottrell