Testing different modes of inheritance

In the main logistic regression practical above we coded 'O blood group status' as a recessive effect of the rs8176719 single base pair deletion. But what if actually the effect is not like that - e.g. maybe it's an additive effect of the SNP (stronger the more copies of the allele you carry?)

Challenge

Create a new variable coding for an additive effect of rs8176719. Then re-fit the regressions using this variable, or with this variable and o_bld_group simultaneously. What's your conclusion?

Hint. Both variables are correlated, so you'll probably find that both appear associated if you test them seperately. But if you include both in the model together, one might 'win out' making it clear which has the strongest association. (Still, including both together is bound to dilute the association signal somewhat - the two predictors are now competing with each other to explain the signal.)

What about a dominant or even heterozygote effect? Are you convinced this is definitely an effect of O blood group?