Fixed effect_Panel

[ajay pasi]

How to include a fixed effect in panel data without a time dimension? Any suggestion or readable?

ajay

[Carlo Lazzaro]

Ajay:
if you actually have a panel dataset, you should have a -timevar- (say, the number of data waves).
That said, it is not mandatory to -xtset- your panel dataset with -timevar- (whereas -panelid- is obviously unavoidable).

[ajay pasi]

Hi carlo, I am dealing with only one wave of the data.

[Carlo Lazzaro]

Ajay:
therefore, you have a cross-sectional study and you should go -regress-.

[ajay pasi]

Thanks for guidlines Carlo. It means

reg depvar indepvars i.deudor, vce(cluster deudor)

will do the trick!

[Carlo Lazzaro]

Ajay:
please note that the number of clusters should be 30-50 to go safely with this type of non-default standard error.
Otherwise, the clustered standard error can easily be misleading.
In addition, if your goal is to corret for heteroskedasticity, please note that, under -regress-, you have to go -robust- instead of -vce(cluster clusterid)-.

[ajay pasi]

Thanks Carlo. great hep, indeed...

[Jeff Wooldridge]

What’s the unit of observation and what’s deudor?

[ajay pasi]

Prof. wooldridge, it is individual-level data (nationally representative). And, deudor is the unique ID for the neighborhood ( data has a total of 340000 deudor IDs)

Actually it is IHDS-II data from India.