search for: 4x2x2x2

...uted, and when data are extra dispersed". In R I coded their C(alpha) test statistic (called TNBI) for interaction for the case where the counts are extra-dispersed, as well as their test for extra-dispersion (called T_a). Using the Quine data set (Quine, 1975) the authors collapse the orginal 4x2x2x2 study design into a 2x4 table and obtained estimates of TNBI=10.36 and T_a=90.81. Using the dispersion estimate from glm.poisson.disp and the estimates for mu I got exactly the same results for TNBI and T_a. This made me happy. Now I thought to try the ML estimates from glm.nb to see if the results...