Using cii proportions with loop

[Celine Robin]

Hello,

I am trying to compute confidence intervals for proportions (number of cases / total) on each observations of a simple database using the ci proportions / cii proportions command.
Here is my code and the database:

input str7 continent cases total
Africa 544 863
America 43 172
Asia 372 734
Oceania 19 25

local contlist Africa America Asia Oceania
    
    foreach continent of local contlist {
     ci proportions total cases if continent == "`continent'"    
    
}

This gives an output with empty values. I have also tried with cii proportions, which was also unsuccesful.

I can get the results one by one, by typing the following command - for Africa for instance - but I would like to have it automatized:

cii proportions 863 544

Does anyone have a idea on how to solve this?

I am using Stata 14.

Many thanks in anticipation

[Nick Cox]

https://www.stata.com/support/faqs/d...iate-commands/ appears to be what you seek.

[David Fisher]

You could also use the meta-analysis package metan for this (via SSC; latest verson v4.06 12oct2022). It leaves behind effect-sizes, standard errors and confidence intervals, and there are options to control how the confidence intervals are calculated (that is, the options of ci proportions). You don't have to perform meta-analysis pooling or draw a forest plot! For instance:

. metan cases total, proportion exact study(continent) nooverall nograph
. list _ES _LCI _UCI

[Celine Robin]

Dear Nick,

Thanks for the response.
I understand that with this do-file I store the output of the cii command in a dataset - which is not exactly what I would like to do. Moreover, with this approach, I would need in any case to type the figures one by one i.e.

 
 run mycii 863 544

(and so on for each continent)

Instead, what I am interested in is simply to display the output of cii from an existing database. I would like to run the cii command for for each observation of my dataset (no need to store the CIs in a new dataset, it's OK for me to have them printed in the result window).

Let me know if this is not clear,

Thanks

[Celine Robin]

Dear David,
Thank you for sharing this package.
The command you proposed did exactly what I needed, thanks,

Best

[Nick Cox]

The idea mentioned in #2 can be adapted to the case of data already in memory.

clear 
input str7 continent cases total
Africa 544 863
America 43 172
Asia 372 734
Oceania 19 25
end 

quietly foreach v in N proportion se lb ub {
    gen `v' = .
}

quietly forval i = 1/`=_N' { 
    cii proportions `=total[`i']' `=cases[`i']' 
    foreach v in N proportion se lb ub {
        replace `v' = r(`v') in `i'
    }
}


    +----------------------------------------------------------------------------+
     | contin~t   cases   total     N   propor~n         se         lb         ub |
     |----------------------------------------------------------------------------|
  1. |   Africa     544     863   863   .6303592   .0164316   .5971731   .6626518 |
  2. |  America      43     172   172        .25   .0330169   .1872156   .3216102 |
  3. |     Asia     372     734   734    .506812   .0184536   .4699971   .5435719 |
  4. |  Oceania      19      25    25        .76   .0854166    .548712   .9064355 |
     +----------------------------------------------------------------------------+

[Celine Robin]

Thank you Nick! Very useful to know this,
Best

[Andrea Discacciati]

See also the user-written command -xcipoibin- (available on github, see below), if you're interested in calculating exact CIs.

https://github.com/anddis/xcipoibin
https://anddis.github.io/xcipoibin/xcipoibin_ex.html

Disclaimer: I'm the author of xcipoibin.

[Celine Robin]

Thanks