R help - Nov 2011 - fitting some form of linear model with bimodal distribution of dependent variable

Hi All,

I have a parameter that is bimodal, and I want to get some sort of linear model
done with it

results = some.linear.function(bimodal.param ~ factor1 + some other stuff,
mydata)

I want to see if factor 1 matters (it has 3 levels, of of which can be taken as
baseline), i.e:

summary(results)

Coefficients:
            Estimate Std. Error t value Pr(>|t|)   
(Intercept) -2.108522   0.936666  -2.251   0.0272 
factor1-1        0.769314   0.273368   2.814   0.0062 
factor1-2        0.149841   0.198976   0.753   0.4537   

[the numbers are made up]

I am not clear which function, if any, could handle such situation. Any
suggestions? I tried glm with family = quasi binomial, but it's obviously
wrong if no other reason that it does not accept the bimodal parameter (it wants
it to be 0s and 1s).

Cheers

Federico





--
Federico C. F. Calboli
Neuroepidemiology and Ageing Research
Imperial College, St. Mary's Campus
Norfolk Place, London W2 1PG

Tel +44 (0)20 75941602   Fax +44 (0)20 75943193

f.calboli [.a.t] imperial.ac.uk
f.calboli [.a.t] gmail.com