I am applying a difference in difference method to a time series with daily [365] observations in both the treated and control groups. To test for parallel trends I was planning on just undertaking a visual inspection of the pre-intervention trends. I am aware that commonly this is done by just looking at the raw data. However, there are 200 observations in each, which provides a messy picture. I have seen some resarch fit simple linear regressions to the pre-intervention treatment and control to provide a clearer picture of whether there are parallel trends - but am worried that this method is largely unjustified within litertaure. Does anyone have any links to papers that have used this approach?
Asked
Active
Viewed 19 times
0
-
1Does this answer your questions? – dimitriy Jan 27 '22 at 18:31
-
Perfect......thanks – j.rahilly Jan 28 '22 at 13:32