R help - Oct 2010 - summary.lm as a post-hoc test?

Let's assume that according to Anova(lm(y~a*b, data=d)) the a:b interaction
is significant, and I would like to know if there are specific combinations of a
and b levels that differ from the control group. Are there any caveats against
simply looking at the p-values in the output generated by
summary(lm(y~a*b,data=d))? The documentation for summary.lm says nothing about
multiple comparison correction, so I assume I need to do that myself, is that
correct?

Thank you.

On 10/30/2010 03:34 PM, Dimitri Liakhovitski wrote:
> You are asking a statistics question, not an R question. R-list people
> never react to such posts.
> Or they give you a nasty reply of the type: "Do you homework first,
> and then ask questions here."
>    
Good point. I thought it was an R question because it was about 
confirming what the output of an R function actually represents, and it 
is not explicitly documented in the help file. I was about 80% sure that 
it is what you just confirmed it to be in your answer, but I didn't want 
to yet again make an assumption about the behavior of this software that 
I'd have to retract later. I am going to hold on to my assumption that 
no multiple comparison corrections were done and it's up to me to do
that.
> Each of the t tests for b:c2, b:c3, and b:c4 test the null hypotheses:
> difference in slopes of a onto b between groups scored c2 (or c3 or
> c4) and the group scored c1 is equal to zero.
Yes! This is precisely the hypothesis I'm testing at the
moment.
> The package I refered to seems to be what you want.
>    
Thank you, I will look into that because most of the time I really do 
need to do pairwise tests of all combinations. Just not this time.