[Anna Chesa]

Hello! I am writing because I would like to plot just a specific number of pre/post periods, but csdid_plot does not allow me. This is my code:

csdid centers_child gdp_pc edu pop nat, ivar(code) time(year) gvar(treatment) agg(event) long2 method(reg)
matrix att = e(b)
matrix V = e(V)
eststo att
esttab att using "$outputpath/figures/CHECKS/centers_child_cs.tex", tex replace
estat event, window(-8,8) plot
csdid_plot, style(rcap) pstyle(p1) graphregion(fcolor(white)) plotregion(fcolor(white)) title("ATT on Centers per Child", color(black))

[FernandoRios]

try
csdid centers_child gdp_pc edu pop nat, ivar(code) time(year) gvar(treatment) agg(event) long2 method(reg)
estat event, window(-8 8)
estat plot

[Anna Chesa]

Thank you Fernando! I have another question: in the command, the option agg(event) is what I should use if I want to aggregate my ATTs as the average effect of participating in the treatment for the group of units that have been exposed to the treatment for exactly e number of periods?
Thank you so much!

[FernandoRios]

exactly

[Tiziana Giuliani]

Good morning FernandoRios

I tried to include in csdid command lagged variables but It came out a message saying that lagged variables cannot be used. Is this related to the fact that csdid uses 2X2 DiD design?

Thanks a lot in advance...
Tiziana

[FernandoRios]

No
its related to the fact that factor notation was not programmed into
alrhough I would suggest to consider what it means to add a lagged variable when estimating attgts

[Julian Spieleder]

CSDID assume a unit is treated only once. then it remains treated.
You can probably analyze the event's to say something about subsequent treatments

Thank you ! In regard to my question and the code then, isn´t it quite easy to adapt the formula for the code in order to handle not just one but also multiple treatments at different times for units? So that the csdid package would be able to handle such matters which would be beneficial. Best, J

[Tiziana Giuliani]

No
its related to the fact that factor notation was not programmed into
alrhough I would suggest to consider what it means to add a lagged variable when estimating attgts

Thank you FernandoRios , I was just wondering whether it could help to increase credibility of the unconfoundness assumption...but may be I am wrong...

Tiziana

[SALINA SIDDIQUA]

Please anyone tell me why this kind of results come applying csdid? Thank you



csdid imr, ivar(cid) time (year) gvar (first_year)
Panel is not balanced
Will use observations with Pair balanced (observed at t0 and t1)
xxxxxxxxxxxxxxxxxxxxxxxxxxxx
Difference-in-difference with Multiple Time Periods

Number of obs = 0
Outcome model : regression adjustment
Treatment model: none
-------------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
--------------------+----------------------------------------------------------------
g2011 |
t_2000_2003.143 | 0 (omitted)
t_2003.143_2006.286 | 0 (omitted)
t_2006.286_2009.429 | 0 (omitted)
t_2009.429_2012.571 | 0 (omitted)
t_2009.429_2015.714 | 0 (omitted)
t_2009.429_2018.857 | 0 (omitted)
t_2009.429_2022 | 0 (omitted)
--------------------+----------------------------------------------------------------
g2014 |
t_2000_2003.143 | 0 (omitted)
t_2003.143_2006.286 | 0 (omitted)
t_2006.286_2009.429 | 0 (omitted)
t_2009.429_2012.571 | 0 (omitted)
t_2012.571_2015.714 | 0 (omitted)
t_2012.571_2018.857 | 0 (omitted)
t_2012.571_2022 | 0 (omitted)
--------------------+----------------------------------------------------------------
g2018 |
t_2000_2003.143 | 0 (omitted)
t_2003.143_2006.286 | 0 (omitted)
t_2006.286_2009.429 | 0 (omitted)
t_2009.429_2012.571 | 0 (omitted)
t_2012.571_2015.714 | 0 (omitted)
t_2015.714_2018.857 | 0 (omitted)
t_2015.714_2022 | 0 (omitted)
--------------------+----------------------------------------------------------------
g2022 |
t_2000_2003.143 | 0 (omitted)
t_2003.143_2006.286 | 0 (omitted)
t_2006.286_2009.429 | 0 (omitted)
t_2009.429_2012.571 | 0 (omitted)
t_2012.571_2015.714 | 0 (omitted)
t_2015.714_2018.857 | 0 (omitted)
t_2018.857_2022 | 0 (omitted)
-------------------------------------------------------------------------------------
Control: Never Treated

See Callaway and Sant'Anna (2021) for details

[FernandoRios]

Few things to try
a) check you have latest versions of csdid and drdid installed in your computer.
b) check gaps in data (tab year gvar)
this will tell me what could be explaining this

[Jon Thomson-Settle]

FernandoRios I am hoping you can help. Apologies if you have gotten this question before.

I have an unbalanced panel of house sales. They are organised by transaction. I have created treatment through spatial matching (e.g. 10km radius). I have created first_treat but when I do: tab year first_treat it looks like this. How do I correct this?

Year	first_treat
0	1993	1994	1995
1991	10913	0	0	0
1992	13663	0	0	0
1993	8862	2980	0	0
1994	6328	0	5616	0
1995	4191	0	0	4473

Second, for ivar in the formula, I can either use the house identifier or transaction sale identifier(first column). If I use the house identifier, they are typically traded infrequently. I would love to use the suburb identifier but there are too many of those transactions per year and the formula does not allow for that.

Would you suggest I use any other model?

Sample of the data
Listed 3 out of 371370 observations

uniqueidentifier	saleprice	year	houseid	treated_5km	first_treat
26306	860000	2021	106938	1	2021
3215	1700000	2015	255824	0	0
3679	1635000	2024	153978	1	2024