[R] Help with Bartlett's test on linear model

[Emmanuel Charpentier]

Stephanie Bernard a écrit :
> Hi all,
> I would like to test the homoegeneity of variances between several
> linear model for some analysis of covariance. It seems that the
> Bartlett's test is a good test to use but I am having problem using with
> linear model and I cannot find any examples on the internet. There are
> some examples for comparisons of variances but not linear models.
> If I take the hellung data set, which is the example in Dalgaard's book.
> I know var.test works fine but I want to learn how to use the Bartlett's
> test.
>> hellung$glucose <- factor(hellung$glucose, labels=c("Yes","No"))
>> attach(hellung)
>> tethym.gluc <- hellung[glucose=="Yes",]
>> tethym.nogluc <- hellung[glucose=="No",]
>> lm.nogluc <- lm(log10(diameter)~log10(conc), data=tethym.nogluc)
>> lm.gluc <- lm(log10(diameter)~log10(conc), data=tethym.gluc)
> I guess I have two questions. 1) How to use bartlett.test with linear
> model (using the model above)

Please look at ?bartlett.test. This generic has methods both for
formulae (formulas ?) and for objects inheriting from the class "lm".
Therefore you might call :
bartlett.test(nogluc)
and
bartlett.test(gluc)
or (equivalently, I think)
bartlett.test(log10(diameter)~log10(conc), data=tethym.nogluc)
etc ... ad nauseam.

and 2) how to test for homogeneity of
> variances of linear models when there are more than two groups.


As the "?bartlett.test" will tell you, this function "[p]erforms
Bartlett's test of the null that the variances in each of the groups
(samples) are the same", which is exactly what you aim at...

HTH

					Emmanuel Charpentier