unbalanced panel - Im–Pesaran–Shin (IPS) error r(498)

[Katharina Koe]

Hi all,

I am currently working on a panel data set that contains:
- N: 40.000
- T: 4
- observations: 50.000

I am trying to determine if the stationarity assumption holds.
Since my panel is unbalanced, my options are limited to the Im–Pesaran–Shin (IPS) or Fisher-type tests.
Fisher-type assumes large T, which is not the case with my data, therefore I prefer the IPS test.

Stata gives error code r(498) that I have gaps in my data.
I have observations for every time period, but I do not have an observation for every household in every time period (my panel is unbalanced). But then what are gaps in the data?

Could someone help by exlaining this to me?
Is there a solution to the problem? Another test maybe?

Many thanks in advance!
Katharina

. xtset newid timevar
panel variable: newid (unbalanced)
time variable: timevar, 1 to 4, but with gaps
delta: 1 unit

. xtunitroot ips alcohol, trend
Im-Pesaran-Shin test cannot have gaps in data
r(498);

. tab timevar

    timevar |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |      8,765       16.80       16.80
          2 |     12,278       23.54       40.34
          3 |     13,266       25.43       65.78
          4 |     17,850       34.22      100.00
------------+-----------------------------------
      Total |     52,159      100.00

[Phil Bromiley]

The gaps in the data are " I do not have an observation for every household in every time period" - you have a gap for those panels. I'm afraid I can't help on your substantive question.

[Katharina Koe]

Hi Phil,
thank you for looking into that.
The gaps (not every HH in every time period) makes my panel unbalanced, right?
I dont neccessarily need to use IPS.
If you knew of any other method, this would be very hepful, too. Maybe you have an idea for an alternative way...?

Thanks!

[Jenny Smart]

I was able to use the xtunitroot fisher commands when I came across the same error running the xtunitroot ips command with my (similarly unbalanced) data:

eg.

xtunitroot fisher wheat_yld, dfuller lags(1) demean
xtunitroot fisher wheat_yld, dfuller lags(1) demean trend

xtunitroot fisher wheat_yld, pperron lags(1) demean
xtunitroot fisher wheat_yld, pperron lags(1) demean trend

https://www.statalist.org/forums/for...code-not-found