Am 01.02.2014 01:40, schrieb Moses Nwanji:
> Dear Sir/Madam,
>
> I am sending this email regarding a query I had about the code needed
> for a 'rolling regression' in gretl.
>
> I am currently trying to analyse the unemployment-output relationship,
> and I plan to use the rolling regression approach to test whether the
> relationship is stable over time (estimated coeffcients). However, the
> issue I have is with the coding, which I have no clue about.
>
Well you will need to learn a little about gretl scripting with the
hansl language.
The general pattern of a rolling regression would probably look like this:
<hansl>
smpl 1970:1 1990:4
loop 100 # some high enough number
ols lhs const rhs # estimate
smpl +1 +1 # move the startobs and the endobs
if $t2 > myfinalobs # $t2 is the sample end number
break # quit the loop
endif
endloop
</hansl>
hth,
sven
