Calculating ratio with confidence interval

[Rana Zahroh]

Dear All,

Do you perhaps have any advice on how to calculate the ratio with 95% confidence interval from the table that is produced from svy, over(exposure): proportion outcome? What kind of command I could use on this?

Here is the result that I got from this command:

. svy, over(birth_place): proportion v401 if bidx==1 & b5==1 
(running proportion on estimation sample)

Survey: Proportion estimation

Number of strata =      67               Number of obs   =     15,028
Number of PSUs   =   1,967               Population size = 14,740.883
                                         Design df       =      1,900

---------------------------------------------------------------------
                    |             Linearized            Logit
                    | Proportion   Std. Err.     [95% Conf. Interval]
--------------------+------------------------------------------------
   v401@birth_place |
  no public sector  |   .7844617   .0079373      .7684879    .7996208
 no private sector  |   .7804379   .0073014      .7657834    .7944217
          no other  |          1          .             .           .
 yes public sector  |   .2155383   .0079373      .2003792    .2315121
yes private sector  |   .2195621   .0073014      .2055783    .2342166
         yes other  |          0  (no observations)
---------------------------------------------------------------------

I would like to calculate the ratio of yes public sector and yes private sector. I have run the following commands to calculate the non linear combination (tried different combinations inspired from website to make it work but failed):

nlcom yes public sector/yes private sector
nlcom [v401@birth_place]yes public sector/[v401@birth_place]yes private sector
nlcom _b[yes public sector]/_b[yes private sector]
nlcom [birth_place]yes public sector/[birth_place]yes private sector
nlcom _b[birth_place:0]/_b[birth_place:1]

Is there something wrong with the commands or did I miss a crucial step prior calculating the nlcom? Appreciate your help on this. Thank you.

[David Radwin]

Have you tried svy: ratio?

Also, you may want to change your syntax to

svy, over(birth_place) subpop(if bidx==1 & b5==1): proportion v401

(The code above is untested.) For an explanation of why subpop() is usually preferable to if with svy estimation, see https://stats.idre.ucla.edu/stata/fa...data-in-stata/ or help svy.

[Rana Zahroh]

Dear David,

Thank you so much for your help! I have found the issue on this - I tried my codes on my friend's Stata 15.l and the codes run very well - yet this does not work on my Stata 16 package. I posted a question on this also here: https://www.statalist.org/forums/for...-16-any-advice. If you know what this issue would be, I would really appreciate if you can give me suggestion. Thank you once again David!

Best,
Rana