R-help,
I am interested in estimating the effect of a treatment (2 levels) on a
response. I've used a randomized blocked experiment (5 blocks). I run
the full model, let's say that it is...
lm1 <- lm(resp ~ treat + block)
...and find that there are no significant block effects. Now with
"treatment" contrasts, the parameter estimate for the non-reference
treatment group can be interpreted as the mean treatment effect
(difference in mean response between the two treatments) in the
reference block only. Is that correct?
I know that this mean is not significantly different from the mean
effect in the other four non-reference blocks (because the block effect
was not significant), but I also know that the mean effect of the
treatment across all blocks is not equal to the mean in the reference
block. In other words, if I run the model without the blocking factor...
lm2 <- lm(resp ~ treat)
I'll get a slightly different effect size, and a different (larger)
estimate of the uncertainty about that effect.
What I really want is the effect size estimate from the second model,
with an estimate of uncertainty for that effect from the first model. Is
that correct?
Is this a situation in which one might modify the contrast matrix for
the block factor? Can you keep block in the model but estimate the
parameter of interest averaged across all blocks?
Any help or clarification on this issue would be appreciated.
Thanks,
Mark
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
