How to add the mean of outcome variable to esttab output after using sdid

[Wenhan Yan]

I am trying to add the information about the mean of the outcome variable after using sdid, so I use

estadd ysumm

to do so. But it seems like sdid did not store ysumm after iteration. So I wonder how can I add the output mean in esttab. Also is there a way to add the Y mean in the graph? Following is the store list return after I run sdid:

scalars:
            e(N_clust) =  49
               e(reps) =  50
                e(ATT) =  -.0819336
                 e(se) =  .0105743
                  e(N) =  2352

macros:
                e(cmd) : "sdid"
              e(title) : "Synthetic Difference-in-Differences"
                e(vce) : "bootstrap"
             e(depvar) : "r_wp"
           e(clustvar) : "state"
            e(cmdline) : "sdid crime_rate state date treat, vce(bootstrap) covariates(pop_1000 r_black r_notwhite unemploymentrate income_per_capita r_HS r_college lgdp tdate tst.."
         e(properties) : "b V"

matrices:
                  e(b) :  1 x 1
                  e(V) :  1 x 1
         e(difference) :  48 x 2
             e(series) :  48 x 3
               e(beta) :  11 x 1
           e(adoption) :  1 x 1
              e(omega) :  48 x 2
             e(lambda) :  49 x 2
                e(tau) :  1 x 3

And here is an example similar to what I am using:

eststo sdid1: sdid crime_rate state date treat, vce(bootstrap) covariates(pop_1000 r_black r_notwhite unemploymentrate income_per_capita r_HS r_college lgdp tdate tstate, optimized) seed(1234) graph g1on 

esttab sdid1  using "tables/Crime rate.tex", title("Crime rate) starlevel ("*" 0.10 "**" 0.05 "***" 0.01) b(%-9.3f) se(%-9.3f) mtitles( "Drug Crime"  ) replace

Crime Rate
----------------------------
                      (1)   
               Crime Rate   
----------------------------
treat              -0.082***
                  (0.011)   
----------------------------
N                    2352   
----------------------------
Standard errors in parentheses
* p<0.10, ** p<0.05, *** p<0.01

but I want:

Crime Rate
-----------------------------------------------------
                                    (1)   
                                Crime Rate   
------------------------------------------------------
treat                             -0.082***
                                  (0.011)   
Crime rate average                  ...
                                   (...)
------------------------------------------------------
N                                 2352   
------------------------------------------------------
Standard errors in parentheses
* p<0.10, ** p<0.05, *** p<0.01

[David Beheshti]

Hi Wenhan,

Not sure about the graph, but you should be able to adapt the following code to add the mean of the outcome to your table:

clear
sysuse auto

eststo m1: reg price mpg
qui summ price
local mu : di %5.2f r(mean)
estadd local mu `mu'

esttab m1, scalars("mu Mean")

[Andrew Musau]

estout is from SSC, as you are asked to explain (FAQ Advice #12). Two issues:

i. You are not saving the statistic with the regression estimates.
ii. Statistics in esttab appear below the vertical line separating coefficients. What you are saving is a statistic.

eststo sdid1: sdid crime_rate state date treat, vce(bootstrap) covariates(pop_1000 r_black r_notwhite unemploymentrate income_per_capita r_HS r_college lgdp tdate tstate, optimized) seed(1234) graph g1on 

sum crime_rate if e(sample)
estadd scalar ysumm= r(mean): sdid1
esttab sdid1 using "tables/Crime rate.tex", title("Crime rate) stats(N ysumm) starlevel ("*" 0.10 "**" 0.05 "***" 0.01) b(%-9.3f) se(%-9.3f) mtitles( "Drug Crime" ) replace

You can label the statistics as you wish.

help esttab

[Wenhan Yan]

estout is from SSC, as you are asked to explain (FAQ Advice #12). Two issues:

i. You are not saving the statistic with the regression estimates.
ii. Statistics in esttab appear below the vertical line separating coefficients. What you are saving is a statistic.

eststo sdid1: sdid crime_rate state date treat, vce(bootstrap) covariates(pop_1000 r_black r_notwhite unemploymentrate income_per_capita r_HS r_college lgdp tdate tstate, optimized) seed(1234) graph g1on

sum crime_rate if e(sample)
estadd scalar ysumm= r(mean): sdid1
esttab sdid1 using "tables/Crime rate.tex", title("Crime rate) stats(N ysumm) starlevel ("*" 0.10 "**" 0.05 "***" 0.01) b(%-9.3f) se(%-9.3f) mtitles( "Drug Crime" ) replace

You can label the statistics as you wish.

help esttab

Hello Andrew,

Thank you for the quick and accurate respond, may I ask if I want to add this statistic into my graph, should I need to list the statisitc first using

ereturn list

and then use yline?

[Andrew Musau]

You could hold it in a local macro or refer to the result stored by summarize if not overwritten, then use the -text()- option to place it in the graph.

webuse grunfeld, clear
sum mvalue if company==1
line mvalue year if company==1, xtitle("") text(5500 1945 "{it: Mean} = `:di %3.2f `r(mean)'' ")

[Wenhan Yan]

estout is from SSC, as you are asked to explain (FAQ Advice #12). Two issues:

i. You are not saving the statistic with the regression estimates.
ii. Statistics in esttab appear below the vertical line separating coefficients. What you are saving is a statistic.

eststo sdid1: sdid crime_rate state date treat, vce(bootstrap) covariates(pop_1000 r_black r_notwhite unemploymentrate income_per_capita r_HS r_college lgdp tdate tstate, optimized) seed(1234) graph g1on

sum crime_rate if e(sample)
estadd scalar ysumm= r(mean): sdid1
esttab sdid1 using "tables/Crime rate.tex", title("Crime rate) stats(N ysumm) starlevel ("*" 0.10 "**" 0.05 "***" 0.01) b(%-9.3f) se(%-9.3f) mtitles( "Drug Crime" ) replace

You can label the statistics as you wish.

help esttab

Hello Andrew,

I just realized that the result the "sdid" store did not have "e(sample)", should I just use

 sum crime_rate  

estadd scalar ysumm= r(mean): sdid1

Since if I include "e(sample)", nothing to return

[Andrew Musau]

This will work only if

count if !missing(crime_rate)
assert r(N)==e(N)

holds. Otherwise, the mean value will be off.

[Wenhan Yan]

This will work only if

count if !missing(crime_rate)
assert r(N)==e(N)

holds. Otherwise, the mean value will be off.

So what if I use something like this:

 
  sum crime_rate if !missing(crime_rate)

I mean now if I only calcualte those who where used in the analysis

[Andrew Musau]

That won't necessarily work because the estimation sample depends on missing values in the RHS variables as well. See the definition of listwise deletion. So the assert condition in #6 must first hold for

sum crime_rate if !missing(crime_rate)

to be valid.

[Wenhan Yan]

That won't necessarily work because the estimation sample depends on missing values in the RHS variables as well. See the definition of listwise deletion. So the assert condition in #6 must first hold for



to be valid.

I see thank you for the explanation, luckly the variables I am using for LHS and RHS does not have missing value.