Confusing about interaction term and covariate pattern in logistic regression

[Tanarat Muangmool]

Hi everyone,

I had 3 variables in my data as following;
1. HIV testing (y) --> coded 0 for no, 1 for yes
2. Gender (gender) --> coded 0 for Male, 1 for Female
3. Age group (agegr) --> coded 0 for <=25 yrs, 1 for >25 yrs.

I tried to find the factors associated with HIV testing.
I used the logistic model to identify factors and interaction term;

Case1: Model with main effect and interaction term

logit y i.gender i.agegr gender#agegr
predict y     // find probability in each covariate pattern
predict n, n     // identify covariate pattern

Case2: Model with only the main effect

logit y i.gender i.agegr
predict y2
predict n2, n

After I ran the commands already, I was wondering that why both cases had the same covariate pattern?
So, Case1 and Case2 had 4 patterns.

Gender	Agegr	Pattern no. (n)
0	0	1
0	1	2
1	1	3
1	1	4

In Case1, Why the program did not include an interaction term for covariate pattern?

Example;
In my imagination of Case1;
Should I have 16 patterns?

Gender	Agegr	Interaction of Gender and Agegr	Pattern no.
0	0	0 (Gender=0, Agegr=0)	1
0	0	1 (Gender=0, Agegr=1)	2
0	0	2 (Gender=1, Agegr=0)	3
0	0	3 (Gender=1, Agege=1)	4
0	1	0 (Gender=0, Agegr=0)	5
0	1	1 (Gender=0, Agegr=1)	6
0	1	2 (Gender=1, Agegr=0)	7
0	1	3 (Gender=1, Agege=1)	8
1	0	0 (Gender=0, Agegr=0)	9
1	0	1 (Gender=0, Agegr=1)	10
1	0	2 (Gender=1, Agegr=0)	11
1	0	3 (Gender=1, Agege=1)	12
1	1	0 (Gender=0, Agegr=0)	13
1	1	1 (Gender=0, Agegr=1)	14
1	1	2 (Gender=1, Agegr=0)	15
1	1	3 (Gender=1, Agege=1)	16

Apologize for my mistake. This is my first post for Statalist.

Thank you in advance,
Mink

[Clyde Schechter]

No. You are misunderstanding what gender#age does. It is the product of gender and age. So, it takes on only two values. It is 1 when gender = 1 (male) and age = 1 (> 25 years) and is 0 for all other combinations of gender and age.

[Tanarat Muangmool]

I apologize for my mistake, thank you Clyde. I will correct it.