Compute a difference rate between two variables

[Michael Duarte Goncalves]

Hello everyone,

I have to use the variables lonely and ca_couple, calculate the proportions of individuals who feel lonely some of the time or often separately by whether they are living with a partner or not.

Then, I need to calculate the difference in the rate of "loneliness" between those two groups.

• lonely takes 1 if the individual feels lonely sometimes or often. And 0 if the individual hardly ever or never feels lonely.
• ca_couple takes 1 if living in couple, and 0 otherwise.

I tried the follwowing code:

tabulate lonely ca_couple, row col
local p_lonely_couple = 100*r(row1col1)/r(row1col2)
local p_lonely_single = 100*r(row2col1)/r(row2col2)
local diff_lonely = `p_lonely_single' - `p_lonely_couple'
display "Difference in the rate of loneliness: " %5.2f `diff_lonely'

Is a better way to do that? Thank you in advance. I obtain that in stata:

. local p_lonely_couple = 100*r(row1col1)/r(row1col2)

. local p_lonely_single = 100*r(row2col1)/r(row2col2)

. local diff_lonely = `p_lonely_single' - `p_lonely_couple'

. display "Difference in the rate of loneliness: " %5.2f `diff_lonely'
Difference in the rate of loneliness:     .

Best,

Michael

[Joseph Coveney]

. . . I need to calculate the difference in the rate of "loneliness" between those two groups.
. . . Is a better way to do that?

regress lonely i.ca_couple

The regression coefficient is the difference in proportions of lonely between the two categories of couple.

If you want some kind of test statistic for it and if your sample size is more than just a handful, then try something like the following.

glm lonely i.ca_couple, family(binomial) link(identity)

[Michael Duarte Goncalves]

Hello Joseph Coveney ,

Thanks a lot!

Michael

[Nick Cox]

Reverting to #1 r(row1col1) and its siblings are guesses at what you would like to exist. It is not an error to refer to an r-class result that is not defined, but the result is missing, which contaminates all later calculations. Hence the code in #1 did not evoke any syntax errors, but that is why the code did not work.

It's rare (although not impossible) that r-class results that are defined are not documented. So, whatever is not documented is unlikely to exist.

[Michael Duarte Goncalves]

Thank you for the explanation Nick.
All is clear now.

Michael