FAQ/Rss - CBU statistics Wiki

Revision 13 as of 2008-10-14 14:21:41

Clear message
location: FAQ / Rss

Sums of Squares used by R in lm, aov and lmer

R calculates sequential or Type I SS by default when using lm, lmer and aov as opposed to the type III SS (default in SPSS) which partials out effects of other terms in the model. A hierarchical approach is suggested which gives the same results for both sums of squares and only looks at the highest order terms in a model. See [http://tolstoy.newcastle.edu.au/R/help/05/04/2981.html here].

Type III SS [http://tolstoy.newcastle.edu.au/R/help/05/08/9560.html may be obtained] using terms of form

fit <- anova(lm(y~x+z)) 
anova(fit, ssType=3) 

and

model.2 <- lme(y~a + b, random=~1|group)
anova(model.2,type="marginal")

for lm, aov and lme. The Anova() procedure can be used to obtain type 3 sums of squares for lmer in R. The Anova () procedure is available by downloading the car extension package from [http://cran.r-project.org/ here.]