Interrupted time series - negative "constant"

[Roberto Vidri]

I'm trying to run a Interrupted time series analysis for single group using the "ITSA" package

Code:

tsset num_days

itsa Total_cases, single trperiod(84) lag(11) posttrend figure replace
actest, lags(12)

The model runs, but my "constant" is a negative number.

I'm not sure how to interpret this, I don't think it should be a negative number, given the data.

Is there an explanation for this?

[Sven-Kristjan Bormann]

The negative constant results from estimating a linear time-trend. If you look at the graph then this negative just the starting point for the time trend before the interruption. So, it is nothing to interpret or worry about. It's just a statistical artefact. If you had fitted a polynomial trend then the constant would be probably around 0 or positive.

[Nick Cox]

More discussion if needed at https://www.stata-journal.com/articl...article=st0394

[Roberto Vidri]

Thank you for your responses!