ZINB main and interaction effects

[Diane Sawyer]

I have a data set on the number of times a individuals fall sick in any given year. I have demographics variables education (= 1 if college degree obtained, 0 otherwise) and income (=1 if high income group, 0 otherwise. Due to the presence of a high number of zero observations and over dispersion, I use a zero inflated negative binomial model using the code:

 
 zinb count_sick i.income##i.education, inflate (c.age)

Count_sick	Coef.	P>|z|
1.income	-0.26	0.028
1.education	-0.07	0
income##education 1 1	0.12	0.027
_cons	-0.5	0.56

I want to calculate the expected number of times that individuals in the following groups fall sick:
Group 1 - Education = 0, Income = 0
Group 2 - Education = 0, Income =1
Group 3 - Education = 1, Income = 0
Group 4 - Education =1, Income = 1.

Could someone help me on how to obtain the values?

Thanks

[Clyde Schechter]

margins income#education

[Diane Sawyer]

Does this take into account the inflation probability? Thanks for your reply.

[Clyde Schechter]

I believe it does. I have not reviewed the formulas specifically used, but the default prediction for -margins- after -zinb- is the expected value of the count outcome. I do not think Stata would call it that if it did not account for the zero-inflation.

[Diane Sawyer]

Thank you so much.