[R] multiple comparisons/tukey kramer

[Tyler Smith]

On 2007-11-23, hadley wickham <h.wickham at gmail.com> wrote:
>>
>> What I need is a reference to the tests implemented in glht, so I can
>> decide which one is appropriate for my data. Sequen, Changepoint et
>> al. may be common terms in some fields, but not in the references I'm
>> working from.
>
> Have you read the vignette:
> http://cran.r-project.org/doc/vignettes/multcomp/multcomp.pdf
> ?

Thanks,

I have, and I've now reread it, as well as looking at the examples in
contrMat. I now understand that Tukey, Dunnett, Sequen et al. are
'convenience' terms, specifying all pair-wise, treatments vs control
etc., contrasts. I worked through the numerical example in Dunnett's
original paper, and also found a worked example of Tukey-Kramer
contrasts, and reproduced the results using 

glht(LM, linfct = mcp(fac = 'Dunnett'))
glht(LM, linfct = mcp(fac = 'Tukey'))

I'm still finding the language very difficult. I was taught these
methods using SS/MS language. Linear models were mentioned in passing,
but the focus was on using (and hopefully understanding) the formulas
for filling in the cells in an anova table. Has the linear model focus
now replaced the 'traditional' approach I learned? 

It's been a very frustrating week or two as I try and come to terms
with material I thought I already knew, so what I'm wondering is if
I'm doomed to repeat this process with every variation on Anova
designs I want to analyze in R unless I take the time to study linear
models. Today's task, for example, is figuring out how to convert
Sokal and Rohlf's two-level nested model II anova into R code, and
it's not going well.

Thanks,

Tyler