I don't find it possible to advise given this information, except
to wonder whether either approach is legitimate. It sounds
as if you can make your best guess at the missing values
_after_ fitting the most appropriate model you can think of.
But doing it the other way, modelling after imputation or
interpolation, will at best overstate the strength of relationship and at worst warp
results by building-in dubious assumptions.
Otherwise put, you are at risk of finding plums in the
dataset that you put there yourself.
Nick
[email protected]
[email protected]
> I have multi Variable Panel dataset believing
> Yit=B0it+B1*X1it+B2*X2it+B3*X3it+.. The problem I need to
> take care is that
> there are missing Y(every 4 years basis ) such that:
>
> ID YR Y X1 X2 X3
> 1 1987 1 3 4 5
> 1 1988 . 4 5 7
> 1 1989 . 7 3 1
> 1 1990 . 8 9 4
> 1 1991 . 7 5 6
> 1 1992 6 7 6 10
>
> 2 1987 3 4 7 9
> 2 1988 . 9 10 14
> 2 1989 . 10 8 20
> 2 1990 . 7 10 14
> 2 1991 . 14 12 10
> 2 1992 2 24 10 16
> (note that this data is made randomly for illustrated purpose)
>
> From the reference and statalist,I may consider "ipolate" or
> "impute" option
> assuming there may be a linear relationship with Ys between years
> I am so glad if anyone give me better suggestions or more
> concern I have to look
> before using those options
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
