Panel Cointegration for data with gaps in the time series

[Justice Djokoto]

Hello,
I have panel data N=19 and T=29 with gaps in the time series. I need to do cointegration. The known Stata tests; xtcointtest and xtwest require data without gaps in the series. I have searched and didn't seem to find a code to handle my data. I am using Stata 15. Any help to find a code to handle my data?
Justice.

[Andrew Musau]

For economic time series, you can fill in missing values by interpolation or extrapolation. Some interpolation methods may be better relative to linear interpolation, depending on the variable. See

help ipolate

and

ssc install mipolate
help mipolate

which has more methods available. Then you just have to state that, e.g., "missing values were filled using linear interpolation and extrapolation".

[Justice Djokoto]

Thank so mcuh, Andrew Musau.
I will try it and revert.
Regards.

[Justice Djokoto]

I got aroud making the panel balanced by filling in the missing values by inter and extrapolation. Thanks.
However, is it the case that no Stata code exists to run cointegration for panel data with gaps (holes) in the time series?

[Andrew Musau]

Have you tried

tsfill, full

before running the command?

[Justice Djokoto]

Sorry for the late response.
No. I did not try it. The gaps were across all the variables. So, I had to compute the interpolation and extrapolations manually.

[Justice Djokoto]

I still need a response to this part of my question: "However, is it the case that no Stata code exists to run cointegration for panel data with gaps (holes) in the time series?"

[Andrew Musau]

I still need a response to this part of my question: "However, is it the case that no Stata code exists to run cointegration for panel data with gaps (holes) in the time series?"

If the gaps are either at the start or end of the series, see https://journals.sagepub.com/doi/pdf...867X1401400312. But the command is for large \(N\) and large \(T\).

[Justice Djokoto]

Thank you, Andrew.