How to treat missing data in a Panel dataset?

[Souryabrata Mohapatra]

Hi,

I’ve a quick question-

I’ve a panel dataset and do have some missing values in the variable column.
I was wondering about using the following command-

• ipolate variable year, gen (ipolate_variable) epolate

Do you think it’s the right way to fill out the missing data?

Thanks,

[Carlo Lazzaro]

Souryabrata:
welocme to this forum.
Yes, -ipolate- can actually do the trick.
However, the substantive issue is the diagnosis of the missing mechanism (basically: are missing values ignorable or not)?

[Souryabrata Mohapatra]

Hi Carlo,

Thank you for your reply.
I find StataList useful.

I've district-level cross-sectional agricultural, demographic & climate data for India.
I attempt to fill the missing data efficiently.
Because at the end I need to do spatial panel analysis.
If I miss too many district-level estimates, that may affect my GeoDa mapping in future, I think.

Mipolate is also useful for population data.
For certain variable like crop-wise harvest price, I find Nnipolate useful as it utilizes neighbouring districts data to fill in the gaps.
But where there's no district-level data, the missing data remains missing.

So I also wonder using 3-year Moving Average, if applicable!
I don't know how to do that in Stata.

As Stata can estimate an unbalanced panel dataset, I attempt to fill out missing data as much as I can.

Kind regards,
Sourya

[Carlo Lazzaro]

Sourya:
you may find -help tssmooth- useful for moving average.
However, whether or not your missing data are ignorable or not should be addressed in your research.