Binary values regression type

[Esmeralda Veen]

Hi hello,

This is my first time posting, so I hope I am doing this right and it is clear for everyone!

I have a dataset where I have a variable that is called charitablegiving. The following responses were possible:

-9 missing
-8 inapplicable
-7 proxy
-2 refused
-1 don't know
1 at least once a week
2 less than weekly, at least monthly
3 less often than once a month
97 other

I created a dummy variable for people who indicated that they give at least once a week using the following code:
gen chargaloaw = charitablegiving == 1 if charitablegiving > 0

I did the same for the variable that indicated whether people had a diploma in higher education:
gen diplhe = diploma == 3 if diploma > 0

I checked if these were correct and the values were all good.

The question I now have is, what type of regression would I use to see the relationship between having a diploma in higher education and donating at least once a week? I believe they are both binary variables now, so do I just use regress or maybe logit?

Thanks in advance!

[Joseph Coveney]

What's "97 Other"? It seems as if response categories 1 through 3 are mutually exclusive and collectively exhaustive.

And category 3 includes not giving to charity at all. So I would recommend something more like an ordered categorical regression (-ologit-, -oprobit- etc.) after recoding the negative values and probably the 97 value to missing values usng -mvdecode-.

Also, the -if diploma > 0- in your second dichotomization seems redundant. If that variable, too, has ordered categories of educational attainment, then maybe use them as-is, too (that is, use factor variable notation), and look at the regression coefficient for that category.

[Joseph Coveney]

Anyway, more to your specific objective, you could try something like the following. It illustrates the use of -mvdecode-. (It assumes that you don't want to include category 97 among the observations submitted to your regression command.)

mvdecode charitablegiving diploma, mv(-9=.m \ -8=.i \ -7=.p \ -2=.r \ -1=.d \ 97=.o)

quietly generate byte chargaloaw = charitablegiving == 1 if !missing(charitablegiving)
quietly generate byte diplhe = diploma == 3 if !missing(diploma)

logit chargaloaw i.diplhe, nolog // For example

// Also you might consider something like
ologit charitablegiving i.diploma, nolog