Dear Ista, > -----Original Message----- > From: istaz...@gmail.com [mailto:istaz...@gmail.com] On Behalf Of Ista Zahn > Sent: March-01-09 6:22 PM > To: r-help@r-project.org > Cc: John Fox > Subject: Understanding Anova (car) output > > Dear professor Fox and R helpers, > I have a quick question about the Anova function in the car package. > When using the default "type II" SS I get results that I don't > understand (see below). > > library(car) > Data <- data.frame(y=rnorm(10), x1=factor(c(rep("a",4), rep("b",6))), > x2 = factor(c(rep("j", 2), rep("k", 3), rep("j", 2), rep("k", 3)))) > Data$c1 <- recode(Data$x1, '"a"=-1;"b"=1', as.factor.result=FALSE) > Data$c2 <- recode(Data$x2, '"j"=-1;"k"=1', as.factor.result=FALSE) > Data$c3 <- Data$c1*Data$c2 > > options(contrasts=c("contr.Sum", "contr.poly")) > > m1 <- lm(y ~ x1*x2, data=Data) > m2 <- lm(y ~ c1 + c2 + c3, data=Data) > > Anova(m1)#Why is this > Anova(m2)#Different from this?
How could Anova() possibly know in m2 that c1 and c2 are marginal to c3 (as in can tell in m1 that x1 and x2 are marginal to x1:x2)? What you're getting, therefore, is equivalent to "type-III" tests: > m1 <- lm(y ~ x1*x2, data=Data) > m2 <- lm(y ~ c1 + c2 + c3, data=Data) > Anova(m1)#Why is this Anova Table (Type II tests) Response: y Sum Sq Df F value Pr(>F) x1 0.3313 1 0.1599 0.7031 x2 0.6800 1 0.3282 0.5875 x1:x2 0.1160 1 0.0560 0.8209 Residuals 12.4321 6 > Anova(m2)#Different from this? Anova Table (Type II tests) Response: y Sum Sq Df F value Pr(>F) c1 0.2715 1 0.1310 0.7298 c2 0.5891 1 0.2843 0.6131 c3 0.1160 1 0.0560 0.8209 Residuals 12.4321 6 > Anova(m1, type=3) Anova Table (Type III tests) Response: y Sum Sq Df F value Pr(>F) (Intercept) 0.5384 1 0.2598 0.6284 x1 0.2715 1 0.1310 0.7298 x2 0.5891 1 0.2843 0.6131 x1:x2 0.1160 1 0.0560 0.8209 Residuals 12.4321 6 > Regards, John > > I would greatly appreciate it if someone can help me understand why > the results differ. > > Thanks, > Ista ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.