Michael.Laviolette at dhhs.state.nh.us
2011-Jul-08 14:26 UTC
[R] Model fit using mcexact from exactLoglinTest
I'm using the mcexact function from the exactLoglinTest package on data
comparing performance of rapid and laboratory tests for detection of H1N1
flu. My setup is as follows:
ridt.res <- c("A-B-", "A+B-", "A-B+")
pcr.res <- c("Negative", "AH3", "B")
xtab <- expand.grid(ridt = ridt.res, pcr = pcr.res)
xtab$counts <- c(271, 68, 7, 2, 74, 0, 3, 0, 9)
xtab$sym.pair <- factor(c(1, 2, 3, 2, 5, 4, 3, 4, 6))
Note that the table is mostly sparse off the main diagonal and that it
contains a symmetric pair of zeros. I'm trying to fit a quasi-symmetry
model with
mcexact(counts ~ ridt + sym.pair, data = xtab)
but I get the following error message:
Error in temp %*% solve(v[(n1 + 1):n, (n1 + 1):n]) %*% t(temp) :
non-conformable arguments
Could you provide some guidance? Thanks.
-M. Laviolette