[R] Assessing standard errors of polynomial contrasts

Daniel Malter daniel at umd.edu
Sun Jul 26 03:24:58 CEST 2009


Hi, using polynomial contrasts for the ordered factors in an experiment
leads to much nicer covariance structure than using treatment contrasts. It
is easy to assess the mean effect for each of the experimental groups.
However, standard errors are provided only for the components of the
orthogonal contrasts. I wonder how to assess the standard errors not of the
components, but of the respective experimental treatment groups as a whole.
If the correlation between the fixed effects is small, can the standard
error be approximated by 
sqrt( sum (SE of components^2))) ?

I know this is a stats rather than an R question, but I thought one of the
many specialists in experiments might be able to help me out quickly on this
or point me to appropriate literature.

Thanks,
Daniel


-----------------------------------------------
"Who has visions should see a doctor," 
Helmut Schmidt, German Chancellor (1974-1982).




More information about the R-help mailing list