VAR error "Repeated time values in sample" even after defining panel dataset

[Kevin Naronha]

I have a panel dataset and I have defined accordingly.

 xtset id week
       panel variable:  id (strongly balanced)
        time variable:  week, 1 to 73
                delta:  1 unit

However, when I use var (vector autoregression) command, I get an error:

. var logtw sumfse3
repeated time values in sample
r(451);

I have made sure that the id and week combination uniquely identifies each entry. Would appreciate any advice. Thanks!

[Kieran McCaul]

I don't do time-series analysis, but your data has an id variable which means that the time variable, week, is repeated by id. Correct?
The command var however appears to expect only a single time-series.

[Kevin Naronha]

Yes, I have multiple observations (week) for each individual (id). In that case, how do I carry out a vector autoregression?

[samialameen]

Hi,
There are two approaches to that; keeping in mind that applying VAR to panel data will complicate the error structure of the estimation, you might look it up in an econometrics book.

1. If you want a formal VAR in panels look for the command <pvar>
2. If you are looking for a crude VAR treatment of a panel data set you need to create a variable that uniquely identify every observation and doesn't cross seams when applying lags to variables. I would do the following as a suggestion;

gen x=id+10000
gen t=week+1000
gen xt=x+t
sort xt
tsset xt

that all to give the data a pseudo time series structure

[Kevin Naronha]

Thanks...I'll try the crude approach to begin with.

[Jack Murphyy]

Hi Kevin, did you manage to solve this issue? I am experiencing the same problem with my data.

[Shubham Sharma]

How to determine the optimum lag length in panel data?