Zoom Logo

Causal Inference Advanced Workshop Day 2 - Shared screen with speaker view
José Lima
37:47
good morning. what is the name of the book? thanks
Scott Cunningham
37:48
Good morning everyone. I am here again to relay questions to Dr. Xu for you.
Rahul Suhag
01:24:14
What's your take on controlling for unobservables using something like inverse mills ratio? If there is selection on unobservables?
Scott Cunningham
01:25:13
I have to take a meeting but will be back soon.
Yiwei Li
01:51:37
Does TWFE works okay under non-staggered situation?
Monika Avila
01:57:52
Could he use the pointer on the computer?
Lev Klarnet
01:58:04
second that!
Kira Vila
01:59:12
yes please!
Amit Mehra
02:00:37
That would be great
Praveen Bhagawan
02:01:05
How does Bacon (2021) test for parallel trend assumption?
Dongryeol Lee
02:03:38
A potential bias with HTE is only problematic when treated observations are treated across different times? (i.e. It won't be a problem in a classic 2x2 design where all of the treatments are treated at the same time)
Seraphim Dempsey
02:06:19
Pointer on computer would be fab!
Grace Eau
02:08:23
Anyway that zoom participants can see what’s in the board?
Farah Mejri
02:09:00
No, I can’t unfortunately
Monika Avila
02:09:41
Thank you! And no problem, it can happen 😃
Lev Klarnet
02:09:43
Can you also please ask to use the computer to point at the slides
Guangying Chen
02:10:30
+1 Thanks!
José Lima
02:11:05
or use the slide editor as yesterday
Grace Eau
02:12:09
Slicer editor use would be great. Please
Laura Nicolae
02:12:16
Where can we read more about adjusting regression for variance weighting or an example of someone doing that in practice?
Kira Vila
02:12:31
+2: when he says "here" or "these guys", we can't tell what he's referring to, which would be fine if it was just sporadically, but it's a regular part of the lecture
Laura Nicolae
02:12:46
I thought last week they said it might be mentioned in Angrist and Pischke but I couldn’t find it
Grace Eau
02:12:46
Agreed
Scott Cunningham
02:22:39
I apologize - I had to have a meeting I couldn't move
Scott Cunningham
02:22:55
I’m back now -- could you please DM me your questions and I can relay them to Dr. Xu if you still have questions?
Lev Klarnet
02:23:20
Can you also please ask him to use the computer to point at the slides
Charles Stoecker
02:23:22
He's still regularly using the laser pointer rather than the mouse cursor.
Scott Cunningham
02:23:29
Yes, I'll ask now
Charles Stoecker
02:23:48
I've emailed Bernie to no avail, so hope you have his cell.
Charles Stoecker
02:26:39
Yes!
Scott Cunningham
02:28:31
Apologies everyone - I had written Bernie but it hadn't gotten to him until now I think
Charles Stoecker
02:30:40
He's using the pointer again.
Kira Vila
02:30:58
If he uses the mouse, it will be visible for everyone.
Monika Avila
02:31:06
Yes
Monika Avila
02:31:12
It is hard to follow
Monika Avila
02:31:41
Thank you again
Yeasung Jeong
02:37:21
For IW, are we assuming the parallel trends between treated units and never-treated units / last-treated units?
Yeasung Jeong
02:38:41
Thank you so much 🙂
Hansen Shao
02:38:45
what if some units receive multiple treatments?
Yeasung Jeong
02:44:35
I have one more question, to use Sun & Abraham or Callaway and Santana, should we test the parallel trend assumptions for all cohorts?
José Lima
02:44:56
What if the notyet treated group can anticipate treatment? maybe because they are in a program with a public schedule of intervention.
Hongyu Fu
02:46:52
@jose I think in Callway & Sant’ana paper, they allow anticipation and make additional assumptions for that.
Lev Klarnet
03:14:05
what is the omitted time period in these event studies?
Xiaoying Yu
03:35:15
why not cluster at the level lower than treatment?
José Lima
04:20:24
we can see,thanks
Charles Stoecker
04:43:14
Hi Bernie, do you have anything to say about ex ante power analyses?
José Lima
05:05:18
I can't see the slides of professor Xu
Scott Cunningham
05:05:36
can anyone see dr. xu's slides?
Yiwei Li
05:07:28
I am seeing a slide titled with "Pretesting and Biases"
Farah Mejri
05:08:21
Yes I do
Farah Mejri
05:09:04
But I can’t hear well anymore!
José Lima
05:19:39
do you need some kind of common support in observables to apply SC method?
Monika Avila
05:37:21
Hello, this presentation is not available in the dropbox. Neither the ones of Wooldridge. Are they available in another repository?
José Lima
05:46:34
this method doesn't require stability of weights in the pretreatment period?
José Lima
05:50:57
this method doesn't require stability of weights in the pretreatment period, in the Hendry's sense of stability?
Rahul Suhag
06:14:29
so, is SDiD always better than Matrix Completion?
Rahul Suhag
06:16:02
yes..
Anurag Pant
06:22:10
If the treatment unit is an outlier in terms of covariates when compared to the control units, does the convex hull imposition make SC more suitable than matching in such cases?
Yeasung Jeong
07:50:10
I have not received a link of the dropbox for the presentation. Could you provide the link so that I can see the presentation slides on my laptop?
José Lima
07:51:38
Where can I find videos of this and previous sessions? Thanks!.
Bernard Black
07:52:23
On the workshop website, under workshop readings, there is a section with all videos. Posted the next day.
José Lima
07:52:52
thanks!
Bernard Black
07:52:59
All slides are on the workshop website, click on workshop readings and then on the link to slides