Is there a reason to limit dummy coding these alleles as 0 or 1 in this regression model example, instead of allowing values of 0,1,or 2?
This is not homework its just an example I found online.
They are testing association between HLA alleles and some binary disease here
in the data frame they have col DQDRa1 and DQDRa2 for haplotypes of HLA gene DQDR. It can be 10 different alleles which they recode into a number of dummy-variables D1, D6, D7, etc., where Di=1 if the subject has at least one
’i’-allele, and =0 otherwise. I don't get why they don't let the Di values go to 2, like with the example at index 460. This individual has diploid for DQDR, but they set D15 col to 1 instead of 2. Why do this?
I think the regression is supposed to look like this is that helps
library(data.frame)
data <- fread("http://www.math.chalmers.se/Stat/Grundutb/CTH/tms121/1011/diabetes.txt")
fit <- glm(Y ~ D6 + D7 + D9 + D10 + D11 + D12 + D13 + D14 + D15 + D99, family="binomial", data=data)
0 answers
No answers yet.
Log in to answer this question.
More posts like this
I assume because they test association for a dominant model, where 0/1 supposedly has the same effect as 1/1.
That would make too much sense :)