Getting proportion of two or more variables with one code

Question

I am analysing a complex survey data. I want to calculate the proportion of smoking among rural men and urban men and alcohol consumption. I used separate code like below. Is it possible to get the same result using one code for urban_rural and alcohol?

svyciprop(~smoking, subset(svs1, urban_rural =='rural'), method = "likelihood")
svyciprop(~smoking, subset(svs1, urban_rural =='urban'), method = "likelihood")
svyciprop(~smoking, subset(svs1, alcohol =='yes'), method = "likelihood")
svyciprop(~smoking, subset(svs1, alcohol =='no'), method = "likelihood")

Can you provide a sample of your dataset? You can read more about that here: https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example — Matt, Jul 15 '20 at 13:05

Thomas Lumley · Accepted Answer · 2020-07-16T05:07:48.213

2

Up to a point, yes, you can use svyby to get the answers for each category of a variable

svyby(~smoking, ~urban_rural, design=svs1, svyciprop, method="likelihood", vartype="ci")
svyby(~smoking, ~alcohol, design=svs1, svyciprop, method="likelihood", vartype="ci")

There isn't any way to do this for a list of variables without some sort of explicit loop or mapping function.

edited Jul 16 '20 at 05:07

answered Jul 15 '20 at 22:28

Thomas Lumley

1,893
5
8

Thank you Thomas. Is it possible to get Confidence Interval too? – Md Shariful Islam Jul 16 '20 at 05:01
You need `vartype="ci"` for `svyby` to return the confidence intervals – Thomas Lumley Jul 16 '20 at 05:08

Getting proportion of two or more variables with one code

1 Answers1