On Thu, 6 Oct 2011, Sven Schreiber wrote:
> Am 06.10.2011 09:45, schrieb Nikos Tsitiridis:
>> Goodmorning,
>>
>> I ran 120 regressions with a script and now I want to re-run them but
>> only with the variables that were found to be *statistically important
>> and positive*. Do you know of any scripting methods to do this
>> automatically so that I won't have to edit 120 regressions by hand?
>> Thank you for your time!
>>
>
> Check out the
> 'omit --auto'
> command.
That command will trim the list of regressors to leave only
"significant" terms, but it will not restrict the list to
variables with positive coefficients. Here's a simple example
script that could serve as a basis for doing what Nikos wants:
<hansl>
open data4-10
list bigX = 2 3 4 5 6 7 8 9
ols 1 0 bigX
matrix b = $coeff
matrix s = $stderr
scalar p = $ncoeff
list trimmedX = null
scalar j = 2
loop foreach i bigX --quiet
if b[j]/s[j] > 2.0 # your criterion here
trimmedX = trimmedX $i
endif
j++
endloop
list trimmedX print
ols 1 0 trimmedX
</hansl>
Another possibility is to look at the code of the "addlist"
function package, which could be modified to suit. To do that,
go to "/File/Function files/On server..." and install the
addlist package. Then "/File/Function files/On local machine",
select addlist, right-click, and choose "View code".
Allin Cottrell
