search

insongkim / PanelMatch

113 stars 34 forks source link

Parallel trend assumption #84

Open ellenjunghyunkim opened 2 years ago

ellenjunghyunkim commented 2 years ago

Thank you for your package!

All my questions are related to parallel trend assumption....!

How can we change the baseline outcome comparison from Y(t-1) to Y(t-2)? This is to address 1-period anticipation before the treatment. And as someone already raised here, it would be wonderful to have the outcome values themselves in addition to seeing the standard difference.

Kind regards

insongkim commented 2 years ago

@ellenjunghyunkim Thanks for using our package.

We plan to include an option for users to conduct placebo tests, e.g., comparing Y{t-1} vs Y{t-2} as you suggested. We will let you know when such functionality becomes available.

ellenjunghyunkim commented 2 years ago

Thank you for your comment!

My question was toward the direction of allowing some anticipatory effect. When the treatment is a choice made by an individual, allowing some anticipatory periods makes sense to me.

To allow k-period anticipation, instead of having Y{t+F} - Y{t-1}, can we change the base pre-treatment year to Y{t-(k+1)} to have Y{t+F} - Y_{t-(k+1)}? This is to compare the future outcomes to the year when we are free from anticipatory effects. It would be nice to have this functionality!

Kind regards,