Logistic regression

[gergana milanova]

Hi,

Just a quick question. My dependent variable is binary and i'm using logistic regression. When calculating the marginal effects at the mean for example and interpreting them - a coefficient on education for example would be -0.02. So this can be interpreted as for example 'an extra year of education reduces the probability of poor health by -0.02.' Could I also say this in percentages to make it more intuitive: 'an extra year of education reduces the probability of poor health by 2%'

Many thanks!

[Marcos Almeida]

Hello Gergana,

Please present the command and the output under CODE delimiters or by installing the SSC dataex, as recommend in the FAQ.

That's the best way to entice helpful replies. Thanks.

[Carlo Lazzaro]

Gergana:
please, follow Marcos' excellent advice.
As an aside, if your coefficient is (say) positive, how can it have a decreasing effect on the dependent variable?

[Maarten Buis]

With marginal effects you can talk about changes in percentagepoints.

[Clyde Schechter]

As an aside, unless your data come from a randomized controlled trial,l you should not be saying that "an extra year of education reduces the probability of poor health" by any amount. You should say "an extra year of education is associated with an X% lower probability of poor health." Avoid using causal language unless your study design actually supports causal inference.