R help - Nov 2011 - quick ANOVA question

Hello,


imagine the following experimental design:

Group - Value

control - 5
control - 6
control - 5

treated1 - 8
treated1 - 9
treated1 - 9

treated2 - 15
treated2 - 16
treated2 - 15

treated3 - 25
treated3 - 30
treated3 - 28


What I like to know is if I apply an ANOVA to this data and choose the
control group as the reference group (using the relevel function) 
what groups exactly are compared? Are only all treated groups 1, 2, 3 tested
against the control group or are all possible combinations tested?

What is compared?

A) Testing: control vs treated1; control vs treated2; control vs treated3;

or

B) Testing: control vs treated1; control vs treated2; control vs treated3;

                treated1 vs treated2; treated1 vs treated3; treated2 vs
treated3;


Cheers





--
View this message in context:
http://r.789695.n4.nabble.com/quick-ANOVA-question-tp4118724p4118724.html
Sent from the R help mailing list archive at Nabble.com.

Un texte encapsul? et encod? dans un jeu de caract?res inconnu a ?t? nettoy?...
Nom : non disponible
URL :
<https://stat.ethz.ch/pipermail/r-help/attachments/20111129/0b40a357/attachment.pl>

> -----Original Message-----
> What I like to know is if I apply an ANOVA to this data and 
> choose the control group as the reference group (using the 
> relevel function) what groups exactly are compared? Are only 
> all treated groups 1, 2, 3 tested against the control group 
> or are all possible combinations tested?
ANOVA doesn't compare one level against another; it compares all levels
against none. *

lm() coefficients and associated p-values can be used to compare treatments to
the 'control' if you use treatment contrasts with 'control' as
the first or 'base' level and there are no other effects in the model.

But see fortune(242). It may not apply to you, but it always applies to me at
least to the extent that I am never quite sure that it doesn't.


S Ellison

*Ish. A more accurate description might be that anova in R compares a model with
two or more levels specified for a factor to a model with no levels specified
other than the default intercept. But that doesn't have as nice a ring to
it, does it?
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:8}}