Question for the STATA staff regarding STATA 17

[Ali Burki]

Hello,
I have a question regarding a new feature for DID regression in STATA 17, which is the estat ptrends test for checking the common trends assumption or more alternatively known as the parallel trends assumption. I have read the new manual on DID regression https://www.stata.com/manuals/tedidregress.pdf. But I want to know exactly what value this test statistic (F) must be for our null hypothesis of linear trends are parallel, i.e., trends of treatment and control group are parallel before treatment occurs(pretreatment time period), to be rejected. I looked at some values in the manual like 39.97, 2.13, etc. I want to know like a "benchmark" value of the test statistic (F) to reject the null hypothesis.
In my personal study, I am getting a test statistic between 7 and 10.

Hoping for a straightforward answer.

[William Lisowski]

Generally, questions for Stata staff should be addressed to Stata Technical Services as described at https://www.stata.com/support/tech-support/ .

As a non-Stata staff member, I'll say that the output of estat ptrends includes the p-value for the test of parallel trends, which can be treated just like any other p-value from a statistical procedure, for example those found in the "P>|t|" column in the output of regress.

Here is sample estat ptrends output taken from the documentation you linked to.

. estat ptrends

Parallel-trends test (pretreatment time period)
H0: Linear trends are parallel

F(1, 999) =   2.13
 Prob > F = 0.1446

The p-value for this example is 0.1466. In the usual terminology, this test is "significant at the 0.1466 level" - which is to say, not particularly significant.

If you want to test the common trends hypothesis at the 95% confidence level, you want a p-value less than (1-0.95) = 0.05, so you do not reject.

[Ali Burki]

Thank you. I have a better understanding of how to use estat ptrends
William Lisowski