Exporting svy: mean and test results

[Varsha Vaishnav]

Hello,

I am doing ttest with survey data for 10 variables. Is there any way to get the output into table format consisting of means of the variables by the group variable, followed by the difference between the means and then the t-statistic with significance stars?

I'm using the following codes:

svy: mean var1 if sample==1 & area==0 & total_int_women1==1, over(Covid19) coeflegend
test _b[[email protected]] - _b[[email protected]]

Thanks.

[Andrew Musau]

svy: mean allows a variable list, so with multiple variables and the same set of -if- restrictions, estimate everything in one go and then store the test results one by one. Here is a reproducible example. When creating the matrix containing the test statistic and the p-value matrix, it is important to name them as the coefficients from svy: mean so as to create the table more efficiently. I use estout from SSC below.

webuse nhanes2f, clear
svyset psuid [pweight=finalwgt], strata(stratid)
estimates clear
eststo m1: svy: mean zinc age weight, over(black)
local i 0
foreach var in zinc age weight{
    local ++i
    test _b[c.`var'@0.black] = _b[c.`var'@1.black]
    mat F`i'= `:di %3.2f `r(F)''
    mat colname F`i'= "c.`var'@1.black"
    mat pval`i'= `r(p)'
    mat colname pval`i'= "c.`var'@1.black"
}
mat F= F1,F2,F3
mat pval= pval1,pval2,pval3
estadd mat F: m1
estadd mat pval: m1
esttab m1, cells("b  F(star pval(pval))") ///
starl(* 0.1 ** 0.05 *** 0.01) nonumb mlab(none) ///
collab(Coef. F-stat.) label

Res.:

. eststo: svy: mean zinc age weight, over(black)
(running mean on estimation sample)

Survey: Mean estimation

Number of strata = 31              Number of obs   =       9,189
Number of PSUs   = 62              Population size = 104,176,071
                                   Design df       =          31

----------------------------------------------------------------
               |             Linearized
               |       Mean   std. err.     [95% conf. interval]
---------------+------------------------------------------------
  c.zinc@black |
    Not Black  |   87.38342   .4985891      86.36654    88.40029
        Black  |   85.08574   1.165209      82.70929     87.4622
               |
   c.age@black |
    Not Black  |   42.45728   .3302272      41.78377    43.13078
        Black  |   40.38561   .7003786      38.95718    41.81404
               |
c.weight@black |
    Not Black  |    71.5006    .182618      71.12815    71.87305
        Black  |   75.34783   .7636796       73.7903    76.90537
----------------------------------------------------------------
(est1 stored)


Adjusted Wald test

 ( 1)  [email protected] - [email protected] = 0

       F(  1,    31) =    4.02
            Prob > F =    0.0537

Adjusted Wald test

 ( 1)  [email protected] - [email protected] = 0

       F(  1,    31) =    7.78
            Prob > F =    0.0090

Adjusted Wald test

 ( 1)  [email protected] - [email protected] = 0

       F(  1,    31) =   23.57
            Prob > F =    0.0000

. esttab m1, cells("b  F(star pval(pval))") ///
> starl(* 0.1 ** 0.05 *** 0.01) nonumb mlab(none) ///
> collab(Coef. F-stat.) label

-------------------------------------------------
                            Coef.      F-stat.  
-------------------------------------------------
[email protected]           87.38342                
[email protected]           85.08574         4.02*  
[email protected]            42.45728                
[email protected]            40.38561         7.78***
[email protected]          71.5006                
[email protected]         75.34783        23.57***
-------------------------------------------------
Observations                 9189                
-------------------------------------------------

[Varsha Vaishnav]

Thank you, Andrew.

[Varsha Vaishnav]

Hello, Andrew

How can I export the above table in the format as shown in the image.

[Andrew Musau]

You can work with a variant of this.To add the observation counts as columns, you need to place them in a matrix. I do not have time to illustrate now. The trick below lies in recoding the -over()- indicator, and keeping one category in the table.

webuse nhanes2f, clear
svyset psuid [pweight=finalwgt], strata(stratid)
estimates clear
eststo m1: svy: mean zinc age weight, over(black)
preserve
replace black= abs(black-1)
eststo m2: svy: mean zinc age weight, over(black)
restore

local i 0
svy: mean zinc age weight, over(black)
foreach var in zinc age weight{
    local ++i
    test _b[c.`var'@0.black] = _b[c.`var'@1.black]
    mat F`i'= `:di %3.2f `r(F)''
    mat colname F`i'= "c.`var'@1.black"
    mat pval`i'= `r(p)'
    mat colname pval`i'= "c.`var'@1.black"
}
mat F= F1,F2,F3
mat pval= pval1,pval2,pval3
estadd mat F: m1
estadd mat pval: m1
esttab m2 m1, cells("b  F(star pval(pval) pattern(0 1))") ///
starl(* 0.1 ** 0.05 *** 0.01) nonumb collab(Mean F-stat.) ///
mlab(Non-black Black)  label drop(*@0.*)

Res.:

. esttab m2 m1, cells("b  F(star pval(pval) pattern(0 1))") ///
> starl(* 0.1 ** 0.05 *** 0.01) nonumb collab(Mean F-stat.) ///
> mlab(Non-black Black)  label drop(*@0.*)

--------------------------------------------------------------
                        Non-black        Black                
                             Mean         Mean      F-stat.   
--------------------------------------------------------------
[email protected]           87.38342     85.08574         4.02*  
[email protected]            42.45728     40.38561         7.78***
[email protected]          71.5006     75.34783        23.57***
--------------------------------------------------------------
Observations                 9189         9189                
--------------------------------------------------------------

[Andrew Musau]

Here is the table with observation counts:

webuse nhanes2f, clear
svyset psuid [pweight=finalwgt], strata(stratid)
estimates clear
eststo m1: svy: mean zinc age weight if black, over(black)
matrix N1= [e(N),e(N),e(N)]
mat colname N1= [email protected] [email protected] [email protected]
estadd mat N1: m1
preserve
replace black= abs(black-1)
eststo m2: svy: mean zinc age weight if black, over(black)
matrix N1= [e(N),e(N),e(N)]
mat colname N1= [email protected] [email protected] [email protected]
estadd mat N1: m2
restore

local i 0
svy: mean zinc age weight, over(black)
foreach var in zinc age weight{
    local ++i
    test _b[c.`var'@0.black] = _b[c.`var'@1.black]
    mat F`i'= `:di %3.2f `r(F)''
    mat colname F`i'= "c.`var'@1.black"
    mat pval`i'= `r(p)'
    mat colname pval`i'= "c.`var'@1.black"
}
mat F= F1,F2,F3
mat pval= pval1,pval2,pval3
estadd mat F: m1
estadd mat pval: m1
esttab m2 m1 m1, noobs nonumb ///
cells("b(pattern(1 1 0)) N1(pattern(1 1 0)) F(star pval(pval) pattern(0 0 1))") ///
starl(* 0.1 ** 0.05 *** 0.01) collab(Mean Obs. F-stat.) mlab(Non-black Black Difference) ///
coeflab([email protected] "Zinc" [email protected] "Age" [email protected] "Weight")

Res.:

. esttab m2 m1 m1, noobs nonumb ///
> cells("b(pattern(1 1 0)) N1(pattern(1 1 0)) F(star pval(pval) pattern(0 0 1))") ///
> starl(* 0.1 ** 0.05 *** 0.01) collab(Mean Obs. F-stat.) mlab(Non-black Black Difference) ///
> coeflab([email protected] "Zinc" [email protected] "Age" [email protected] "Weight")

--------------------------------------------------------------------------------
                Non-black                     Black                Difference   
                     Mean         Obs.         Mean         Obs.      F-stat.   
--------------------------------------------------------------------------------
Zinc             87.38342         8304     85.08574          885         4.02*  
Age              42.45728         8304     40.38561          885         7.78***
Weight            71.5006         8304     75.34783          885        23.57***
--------------------------------------------------------------------------------

[Varsha Vaishnav]

Thank you, Andrew.