statistical significance between two proportions

[Nicolas Rodriguez]

Hello, I have a survey data and two proportions. This proportions measures the participation level in a program for men a women. In the case of women, the percentage is 98,7% and for men, the percentage is 97,1%. So I want to test whether this difference is statistically significant.
Does anyone knows how to test this??
Regards

[Clyde Schechter]

This requires more information than you supply. The difference might or might not be statistically significant depending on how many women and men there were in total. See -help tabulate twoway- with the chi2 or exact options. Or, from SSC, see -help csti-.

[Nicolas Rodriguez]

I have 170 men and 205 women. I thought about using prtest but this command do not stand svy option. It is a complex survey data so I should consider that

[Clyde Schechter]

Make sure your data is correctly -svyset-. Then run -svy: participated, over(sex)-, followed by -proportion, coefl-. Looking at the results of -proportion, coefl-, select the names shown in the Legend column of the output and use the -test- command to see test them for equality. See -help proportion-, -help test-.

[Steve Samuels]

Crossed with Post #4 above.
Your complex survey data came from a multi-stage, clustered design with strata and weights. Therefore, you cannot use a non-survey command like prtest, which assumes observations are independent and equally weighted. You must svyset the data with the survey design information (strata, primary sampling units, weights) Then svy: tabulate will provide a valid test of difference. Be sure to show all code and results between [CODE] and[/CODE] delimiters, explained in FAQ 12.

[Nicolas Rodriguez]

Thank you very much for all your comment. My command and results are as follows:

use BASE_USUARIO_corregida, clear

keep if region==12

*Poblacion en estudio (mayores de 12 años)
gen poblacion_total= (c14_1_1>=12)

*Porcentajes de participacion día tipo según sexo
gen pc_part_dt=p_trab_fscn_dt/poblacion_total

svyset VarUnit [pw=wgt2], strata(VarStrat) singleunit(certainty)

svy: ratio p_trab_fscn_dt/poblacion_total, over(c13_1_1)

--------------------------------------------------------------
             |             Linearized
        Over |      Ratio   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
_ratio_1     |
      Hombre |   .9708393    .013409      .9131451    1.028534
       Mujer |   .9865298   .0027645      .9746352    .9984244
--------------------------------------------------------------

So how can I test whether this difference are significant?

[Clyde Schechter]

test Hombre = Mujer

will do it, I think.

If that does not work, run

ratio, coefl

and then use the names given in the Legend column of the output in the -test- command instead of Hombre and Mujer.