Repeated time values within panel

[Mert Can Duman]

Dear All,

I would like to generate a gravity model including product and I have three-dimensional-panel data. When I try to generate xtset command w.r.t. destination and product, I have an error as repeated time values within panel.

Moreover, I tried to group j and k into id and try to generate xtset as xtset id t but it gives me the same error.

I will appreciate with your precious support.
M.

t	k	j	exp_i	ln_exp_i	exp_j	j_gdp	j_ln_gdp
1996	1205	528	1.854	0.6173455		296.894	5.693377
1996	1206	40	40.291	3.696128	4.597279	2989.77	8.002951
1996	1206	58	26.964	3.294503		27489.6	10.22156
1996	1206	208	15.009	2.70865		3565.71	8.17912
1996	1206	246			2.991172	25783.5	10.15749
1996	1206	251	2417.353	7.790429	2.900212	2687.29	7.896287
1996	1206	276	197.306	5.284756	7.391737	3485.87	8.156472
1996	1206	300	57.139	4.045487		13749.1	9.528729
1996	1206	372	0.748	-0.2903523		2835.9	7.950114
1996	1206	380	456.444	6.123466		2381.65	7.775548
1996	1206	528	219.883	5.393095	7.092826	296.894	5.693377
1996	1206	724	1546.549	7.343781	3.502429	1619.84	7.390085
1996	1206	752	12.957	2.561636		32999	10.40423
1996	1206	826	4.918	1.592902	5.379565	2446.43	7.802385

[Clyde Schechter]

As you don't show the actual command(s) used to "group j and k into id and try to generate xtset as xtset id t" it is impossible to say what you might have done wrong.

Suffice it to say that when I use your example data and run:

egen id = group(k j)
xtset id t

it runs with no error messages.

Now, your example data is rather scanty, and, in particular, it only continues one value of t. I suspect that the problem arises from other observations in your data set that are not shown. I suggest you post back with the actual code you are trying to use, and also with a new example data set that actually reproduces the error you are getting.

One other suggestion: I have never known Stata to be wrong when it reports this duplicate time values error. You appear to believe that the combination of j and k as a panel identifier and t should work properly for -xtset-. If that is correct, then the combination of j, k, and t should uniquely identify observations. So you can find the observations that are causing your problem by running:

duplicates report j k t, gen(flag)
browse if flag

Then you will need to figure out whether these surplus observations for some combinations of j, k, and t are data errors, or whether they are correct observations and you need something more than just j and k to identify panels for your analysis, or whether the data are correct and complete and you really just shouldn't have a time variable in your -xtset- command.