ESTAT PTRENDS with cross-sectional data without two pre-intervention time periods

[Albino Bruno]

Hi to everyone,

I am doing my computations for a study using DD estimation. After I carried our my difference-in-difference estimation with the didregress command, I run the estat trendplots and the estat ptrends. When I run the last command, the software displays me the following error "estat ptrend requires at least two pre-intervention time periods". I think that's given by the nature of my data (attached). Is there anything I can can do to solve this problem?
Plus, since the estat ptrends command was not working, I tried another way to show - at least graphically - that the parallel trends assumption was somehow holding. And I wrote:

collapse (mean) stfeco, by(essround country)
twoway line stfeco essround if country==1 || line stfeco essround if country==0

But this does not really show the before-after situation I am expecting.
How do I obtain something similar to the picture below?

Thank you in advance.

[Joerg Luedicke (StataCorp)]

Your time variable essround is binary, and so it appears that you have only two time points, one before and one after your event of interest. Thus, with just a single point per time period, a trend cannot be plotted (or tested for). As an aside, have a look at the standard errors of your didregress estimate, they are implausibly small. Remember that the cluster-robust variance estimator does not perform well if the number of clusters is small.

Hope this helps,
Joerg