[FernandoRios]

Yes and no....
you need to use option `long2' to get estimates for pretreatment that are comparable to standard Event studies program
csdid y x, ..... long2

Aftert that, csdid_plot will produce a graph with a cap at T-1. But, the gap will be empty.
HTH

[Kal El]

Hmmm, that plot was produced after running the csdid command with the long2 option, but didn't produce a gap at T-1. I guess I'll just create it in excel.

[FernandoRios]

can you show me what you get if you use the following code;

ssc install frause
frause mpdta, clear
csdid lemp, ivar( countyreal) time(year) gvar(first) long2
estat event

For me it produces this:

------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
     Pre_avg |   .0175956   .0169201     1.04   0.298    -.0155672    .0507585
    Post_avg |  -.0772398    .019965    -3.87   0.000    -.1163705   -.0381092
         Tm4 |   .0033064   .0244519     0.14   0.892    -.0446184    .0512311
         Tm3 |   .0250218   .0181189     1.38   0.167    -.0104906    .0605343
         Tm2 |   .0244587   .0142364     1.72   0.086    -.0034441    .0523616
         Tp0 |  -.0199318   .0118264    -1.69   0.092    -.0431111    .0032474
         Tp1 |  -.0509574   .0168935    -3.02   0.003     -.084068   -.0178468
         Tp2 |  -.1372587   .0364357    -3.77   0.000    -.2086713   -.0658461
         Tp3 |  -.1008114   .0343592    -2.93   0.003    -.1681542   -.0334685
------------------------------------------------------------------------------

If you get something else, you may need to update your csdid and drdid files.

[Kal El]

Here's my code:

csdid PQI90r PercentUnemployed PercentBelowPoverty [iweight=att_wt1], ivar(hsanum) t(YEAR) gvar(y_start) method(dripw) long2
estat event

For me it produces this:

ATT by Periods Before and After treatment
Event Study: Dynamic effects
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
     Pre_avg |  -6.700665   45.24749    -0.15   0.882    -95.38412    81.98279
    Post_avg |  -27.61554   24.64424    -1.12   0.262    -75.91736    20.68629
         Tm4 |   -47.4056   78.09454    -0.61   0.544    -200.4681    105.6569
         Tm3 |   5.415585   71.28649     0.08   0.939    -134.3034    145.1345
         Tm2 |   11.29903   44.29723     0.26   0.799    -75.52195       98.12
         Tm1 |   3.888325   23.92913     0.16   0.871     -43.0119    50.78855
         Tp0 |  -34.43216   17.09996    -2.01   0.044    -67.94746   -.9168607
         Tp1 |  -11.13073   25.41266    -0.44   0.661    -60.93862    38.67716
         Tp2 |  -41.11549   28.36162    -1.45   0.147    -96.70324    14.47226
         Tp3 |  -23.78376   60.03834    -0.40   0.692    -141.4567    93.88922
------------------------------------------------------------------------------

[FernandoRios]

Ok try updating csdid
or at least verify its version 1.7 or higher

[Richard Thomas Boylan]

How do I one sided test with csdid?
For instance, suppose I run
use https://friosavila.github.io/playing...rdid/mpdta.dta, clear
csdid lemp lpop , ivar(countyreal) time(year) gvar(first_treat) agg(event)
How do I compute, for instance, then likelihood that Tm3 is less than or equal to -0.2?
I saw https://www.stata.com/support/faqs/s...-coefficients/
However, those instruction require r(df_r) and r(F) which are not saved by csdid.

[FernandoRios]

Just think of the coefficients as a regular regression. Use test or Lincoln to construct the modified coefficient and then evaluate it using z score appropriate for your case

[Richard Thomas Boylan]

Ok, thanks!

[Jad Tamimi]

Hello,

Im also experiencing a similar problem, csdid_plot is not producing a T-1 gap for me, the coefficients are corresponding to -1,-2,-3 periods to treatment when they should be corresponding to -2,-3,-4.

Here is the code I used:

* load the dataset
use "$temp/Master enterprise dataset collapsed at ward year with 1999.dta", clear

* tell csdid what control units to use
replace sezyear=0 if sezyear>= 2013 & sezyear<=2016 

* DiD estimation using notyet option
csdid empend, ivar(wardnew) time(year) gvar(sezyear) long2 notyet

* creates and exports a plot for g2003 estimates using notyet option 
qui: csdid_estat attgt 

csdid_plot, group(2003) title("Group 2003") style(rcap) legend(label(1 "95% CI") label(2 "Pre-Treatment Coefficient") label(3 "95% CI") label(4 "Post-Treatment Coefficient") region(lstyle(foreground)) rows(2) position(6))

Here is the graph (pdf):

2003_c_notyet.pdf

Please could you advise me on how I may fix this? Thanks.

[Jad Tamimi]

When using estat event after my estimation the following comes up, however I'm not doing a graph for this output (in this output the T-1 Gap is produced, but not produced in the group output when graphing)

. estat event
ATT by Periods Before and After treatment
Event Study:Dynamic effects
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
     Pre_avg |  -229.2279   79.55591    -2.88   0.004    -385.1546   -73.30118
    Post_avg |   2456.201   394.1057     6.23   0.000     1683.768    3228.634
         Tm9 |   -203.825   195.1824    -1.04   0.296    -586.3756    178.7256
         Tm8 |  -166.2542   107.5019    -1.55   0.122     -376.954    44.44556
         Tm7 |  -301.1865   129.5201    -2.33   0.020    -555.0412    -47.3319
         Tm6 |  -294.5734   107.8697    -2.73   0.006    -505.9942   -83.15266
         Tm5 |   -289.259   96.66711    -2.99   0.003    -478.7231   -99.79495
         Tm4 |  -239.7188   72.90645    -3.29   0.001    -382.6128    -96.8248
         Tm3 |   -182.408   56.02195    -3.26   0.001     -292.209   -72.60701
         Tm2 |  -156.5981   38.59177    -4.06   0.000    -232.2366   -80.95963
         Tp0 |   163.4253   57.88782     2.82   0.005     49.96722    276.8833
         Tp1 |   401.6088   95.38017     4.21   0.000     214.6671    588.5505
         Tp2 |   650.5233   147.7758     4.40   0.000     360.8881    940.1584
         Tp3 |   984.8339   172.5349     5.71   0.000     646.6717    1322.996
         Tp4 |   1426.346   221.2178     6.45   0.000     992.7673    1859.925
         Tp5 |   2661.896    390.341     6.82   0.000     1896.842     3426.95
         Tp6 |   4068.035   648.1162     6.28   0.000      2797.75    5338.319
         Tp7 |   4065.451   749.7587     5.42   0.000     2595.951    5534.951
         Tp8 |   4867.701   983.8101     4.95   0.000     2939.469    6795.934
         Tp9 |   5272.187   1544.767     3.41   0.001     2244.498    8299.875
------------------------------------------------------------------------------

[FernandoRios]

Oh ok
i think I understand the problem
you are trying to produce an event like plot using a single cohort

i will have to modify the code for that. But so not have a date for the update. Your other alternative is to use csdid2.
please see my personal website and repository and contact me if you need help producing the event like plots

will try to post an example here tomorrow
Fernando

[Jad Tamimi]

Perfect thank you!

[FernandoRios]

Here is the code

net install csdid2, from(https://friosavila.github.io/stpackages)
ssc install frause
frause mpdta, clear
** No need for long2, because its the default
csdid2 lemp, ivar(county) time(year) gvar(first)  
estat event, rgroup(2007) plot

HTH

[Jad Tamimi]

Thank you Fernando, this code solved my issue.

Follow up question: in my previous code for csdid_plot I had plot options which look like this:

csdid_plot, group(2003) title("Group 2003") style(rcap) legend(label(1 "95% CI") label(2 "Pre-Treatment Coefficient") label(3 "95% CI") label(4 "Post-Treatment Coefficient") region(lstyle(foreground)) rows(2) position(6))

How can I implement these options using this approach?

Thanks,
Jad

[Jad Tamimi]

Hello Fernando, please disregard the above message, I was able to figure out how to implement these options by using the following code after the csdid2 estimation:

qui: estat event, rgroup(2007)

csdid2_plot, title("Group 2007") style(rcap) legend(label(1 "95% CI") label(2 "Pre-Treatment Coefficient") label(3 "95% CI") label(4 "Post-Treatment Coefficient") region(lstyle(foreground)) rows(2) position(6))

Thank you so much for your help, I appreciate this greatly!