Difference in Difference

[Pace Pace]

Hi All,

I face a problem when I want to check the estat trendplots there is an issue estat trendplots requires at least two pre-intervention time periods. How do I solve it?
I have two pre intervention time periods June 2022 and Nov 2022 and one time post intervention June 2023

Here is the code that I use for my DiD
xtdidregress (revenue) (did), group(idbanksampah) time(bulan)

Treatment and time information

Time variable: bulan
Control: did = 0
Treatment: did = 1
-----------------------------------
| Control Treatment
-------------+---------------------
Group |
idbanksampah | 40 40
-------------+---------------------
Time |
Minimum | 1 2
Maximum | 1 2
-----------------------------------

Difference-in-differences regression Number of obs = 240
Data type: Longitudinal

(Std. err. adjusted for 80 clusters in idbanksampah)
------------------------------------------------------------------------------
| Robust
revenue | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
ATET |
did |
(1 vs 0) | 407319.3 176161.3 2.31 0.023 56678.99 757959.7
------------------------------------------------------------------------------
Note: ATET estimate adjusted for panel effects and time effects.

[Jared Greathouse]

I don't understand the problem. What error did Stata tell you? Post your dataset using dataex and then show your code, in a code block.



EDIT: by the way, you have a very short time series with gaps. I know DID can work with this but you only have 2 preintervention periods. Why should anyone take the parallel trends assumption seriously in this case when you're only fitting to two preintervention periods that aren't even consecutive to one another?

[Pace Pace]

aa

[Pace Pace]

I don't understand the problem. What error did Stata tell you? Post your dataset using dataex and then show your code, in a code block.



EDIT: by the way, you have a very short time series with gaps. I know DID can work with this but you only have 2 preintervention periods. Why should anyone take the parallel trends assumption seriously in this case when you're only fitting to two preintervention periods that aren't even consecutive to one another?

estat ptrend
estat ptrend requires at least two pre-intervention time periods

estat trendplots
estat trendplots requires at least two pre-intervention time periods

[George Ford]

I'd use the natural log of revenue.