Original LR test includes ll(unconstraint)=ll(a)+ll(b), ll(constraint)=ll(a&b) with a restriction of df=2, chi square=-2( ll(constraint)-ll(unconstraint))
Now modify one value in original model and generates two more final results, in original model have variable w=1, in two new models w=0.5, w=3
now need to test the null hypothesis the best between the three is a=1, how to do LR test?
ll(unconstraint)=max {ll(action&belief)-ll(action)-ll(belief)), w=0.5, 1, 3}
ll(constraint)={ll(action&belief)-ll(action)-ll(belief)), w=1}
with the original ristriction gives a df=2, in this new test df=3
is this right?
dissertation deadline approaching, sos .....