Dear all,
I am working on a heterogeneous panel data of 10 Sub-Saharan African countries from 1980 to 2015 using 3 variables. One of the countries has 2 data-points missing from one of the variables, thus I keep getting the error message when trying to test for panel cointegration using either Pedrini (1999) or Westerlund approaches.
Any ideas on how to overcome this challenge? Or can anyone suggest any other panel cointegration test that can accommodate missing values and the way to go about it....I use Stata13.
Thank you.
Maybe the error correction based cointegration test by Gengenbach (xtectest) can deal with missing values.
Here you find further information:
https://sites.google.com/site/medevecon/code
Thanks so much Margarete, I'll try the code and revert!
...gracias!!!
If the time span after the missing values is not too short, you may also test it for sub-period. For example, if the missing values are for 1981 and 1982, you may test it with "if year>=1982". However, this won't be helpful if the missing values are towards the middle of the time span. Good luck!
Thanks Hassen, the missing value is relating to just a group within the panel and its for 1989 and 1990. Since the code is xtpedroni, any suggestion on how to specify it?
One possible solution is to interpolate [fill in something] for the missing values:
1. reg var year [where "var" is the variable with missing values]
2. predict varhat, xb
3. replace var = varhat if year==1989|year==1990.
4. drop varhat [if you want to]
Then you can run the tests.
Hope this helps!
- Badges
- Science topic