thickening <- c( 454, 249, 349, 249, 408, 322, 312, 347, 523, 268, 411, 285, 364, 283, 266, 290) treat <- c(1,2,3,4, 2,1,4,3, 3,4,1,2, 4,3,2,1) site <- gl(4, 4, 16) class <- gl(4, 1, 16) wey <- as.factor(treat > 2) vol <- treat vol[treat==2] <- vol[treat==3] <- 0 vol[treat==4] <- -1.0 fit <- glm(thickening ~ site + class + wey + vol, family=poisson()) resid <- (thickening - fit$fitted)/sqrt(fit$fitted) X2 <- sum(resid^2) s2 <- X2 / (16-9) summary(fit, dispersion=s2) Coefficients: Estimate Std. Error z value Pr(>|z|) (Intercept) 5.91883 0.01788 331.087 < 2e-16 *** site2 0.09579 0.01746 5.487 4.08e-08 *** site3 0.13343 0.01706 7.819 5.32e-15 *** site4 -0.04814 0.01811 -2.659 0.00785 ** class2 -0.43947 0.01722 -25.526 < 2e-16 *** class3 -0.26785 0.01630 -16.430 < 2e-16 *** class4 -0.39675 0.01700 -23.342 < 2e-16 *** weyTRUE 0.21214 0.01743 12.174 < 2e-16 *** vol 0.20950 0.01244 16.839 < 2e-16 *** --- (Dispersion parameter for poisson family taken to be 0.2014748) Null deviance: 265.3027 on 15 degrees of freedom Residual deviance: 1.4038 on 7 degrees of freedom AIC: 141.52 Number of Fisher Scoring iterations: 3