Displaying 1 result from an estimated 1 matches for "test11d".
Did you mean:
test111
2009 Aug 25
0
comparing tables from replicated data
...879 3850
For the less complex model, the results on 1000 replicates are
"evidently" better than the TT default
table(test11b)
test11b
0 1 2 3 4 5 6 7 8 9 10
279 294 191 114 53 40 17 6 3 2 1
and for the more advanced model I get even more convincing
table(test11d)
test11d
0
1000
Clinically I can say that it is bad to have 1 in 16 misclassified and
therefore judge the complicated model better, but others might disagree.
Also it is not too good that the method here is insensitive to the size
of the changes.
I hope some of you will have remarks on this...