I want to use multivariate logistic regression for disease association (binary outcome) of HLA genes. I want to control effect of coexisting alleles of ethnicity haplotypes. Imagine that I have two genes (#layer1) and each gene has 4 alleles (#layer2) as bellow.
HLA-A
-A1
-A2
-A3
-A4
HLA-B
-B1
-B2
-B3
-B4
The problem is that since each person has 2 copies of allele, #layer2 items are not mutually exclusive and therefore this layer should be multivariate in turn (of course if use genotypes, only #layer1 would be multivariate and therefore #layer2 items are numbered as members of #layer1 variables. I have previously done this work Article Association of FOXP3 gene polymorphisms with risk of preecla...
table 4).I can consider #layer2 as the only layer and number the members of each variable as 0 (neg) and 1 (pos) like the above paper as bellow.
A1
0
1
A2
0
1
...
and separately for HLA-B
...
However, both layers are important to me. Because I want to remove confounding effect of both couple alleles and ethnicity associated haplotypes.
What should I do?
I am not familiar with stata commands. Please answer me via stata menus.
Kind regards!