[R] Implementation of contrasts

[Dennis Fisher]

R 3.2.1
OS X

Colleagues

I am trying to understand ANOVA contrasts and I have encountered some puzzling results.

I have data from five studies with four different combinations of treatments.  I want to make the following comparisons:

Study 1, 3: 	A / B vs. C / D (i.e., mean of A and B vs. mean of C and D)
Study 2: 		A vs. B / C
Study 4: 		A / B / C vs. D
Study 5: 		A vs. B / C / D

With some trial-and-error, I got the following code to yield results matching SAS outputs:
	AB.CD 					<- c(1, 1, -1, -1)/2 
	A.BC 					<- c(1, -1, -1)/2
	ABC.D 					<- c(1, 1, 1, -1)/2
	A.BCD 					<- c(1, -1, -1, -1)/2 
	if (STUDY %in% c(1,3)))		contrasts(DATA$TRT)     <- AB.CD
	if (STUDY == 2) 			contrasts(DATA$TRT)     <- A.BC
	if (STUDY == 4)			contrasts(DATA$TRT)     <- ABC.D
	if (STUDY == 5)			contrasts(DATA$TRT)     <- A.BCD

AB.CD makes sense to me — take one-half of each of A and B compare to negative one-half of C and D (the contrasts add to zero).
However, I don’t understand how the other contrasts are written (i.e., they don’t add to zero).  For example, I tried:
	A.BC					<- c(2, -1, -1) / 2
	ABC.D 					<- c(1, 1, 1, -3)/3
without success (they yielded results markedly different from SAS)

I have searched the web extensively but the explanations of contrasts in R are not particularly understandable.  Can anyone help me understand the specifics of this situation?  Thanks in advance.

Dennis

Dennis Fisher MD
P < (The "P Less Than" Company)
Phone: 1-866-PLessThan (1-866-753-7784)
Fax: 1-866-PLessThan (1-866-753-7784)
www.PLessThan.com