Comparing household income between two groups

[Clara Moller]

Hi all,

So I have a dataset where my two dependent variables are 1) underwent surgery (900 patients) and 2) background population (2700 people), then I want to compare their household income with logistic regression in my univariate analysis, to see if there is any difference. But I can't really figure out which way would be the best to do it. I have made quintiles to show how their income is distributed (mean, median, max, min) - should I use mean in each quintile to compare them?
Tried to figure out what others studies do and why they do it, but it seems like it's very different.

Thank you!

[Ken Chui]

Since it says "I have made quintiles," do you have the original income in dollars? If you have that, why not use that instead of using quintile summaries?

[Clara Moller]

Yes, I have the original income. Do you mean that I should just compare them as a continous variable?

[Clara Moller]

My outcome needs to be in odds ratio, so I need something to compare it to. If I use my example above with quintiles, I would compare it to the 5th quintile (my reference).

[Felix Bittmann]

I am a bit confused by the approach. To test whether the relevant variable (income) differs between the two groups (surgery yes or no / 0 and 1) you can conduct a simple t-test. Use the original and continuous income variable, like

ttest income, by(surgery)

However, if you want to test whether the income situation affects the treatment status, you can do that with the logistic regression like:

logit surgery c.income, or