# Linear statistical model: inference about interaction coefficients, two-factor

1. Nov 26, 2008

### solar42

1. The problem statement, all variables and given/known data
I have measurements of some response of a gene, and two factors: the gene, g=1...G and whether the patient/subject has a certain disease, t=1,2.

the full model is
$$y_{gtk}=\mu+\alpha_g+\beta_t +(\alpha\beta)_{gt}+\epsilon_{gtk}$$

I know that to see if genes have any connection at all with the disease, I just fit the reduced model without the $$(\alpha\beta)_{gt}$$ interaction and compare the two, but if I want to see if, say, gene number g=25 has anything to do with the disease... I know that the null hypothesis is $$H_0 : (\alpha\beta)_{25,t}, \textrm{ equal for all } t$$, but how do I test this hypothesis? I am confused at what to do when I can't drop the whole factor and compare.
I don't want to know how to do this in R or something, but how to do it by hand.

2. Relevant equations

3. The attempt at a solution
Well, there are only two levels t=1,2 , so we can basically test $$H_0 : (\alpha\beta)_{25,1}-(\alpha\beta)_{25,2}=0$$, but how!?
1. The problem statement, all variables and given/known data

2. Relevant equations

3. The attempt at a solution

Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Can you offer guidance or do you also need help?