Hi Arthur
In continuation of my earlier question regarding SE of difference (which is
attached below), could you please clarify on the calculation of overall SED
in the .pvs file
for example:
n1, n2, n3 are the number of observations of 3 levels of the fixed effect
and n1* n2 * n3
mse = mean square error
then over all SED = (mse/3 * (1/n1 + 1/n2 + 1/n3))^0.5
but asreml seem to be calculating SED (if I am right!!!!) as:
(mse * (1/n1 + 1/n2 + 1/n3))^0.5
Any comments are appreciated
thanks
Kishore
Hi,
>
> Can the overall standard error of difference (given in the .pvs file for
> predicted means) be used for pair wise comparison of means to determine
the
> significance of the differences between the levels of a fixed effect and
if
> so how?? Is there a need to estimate SE of difference for each pair under
> comparison by taking into consideration the number of records for each
level.
>
> Thanks in advance
>
> Kishore
>
If predicted values have more or less the same accuracy, then the
average SED is used to construct a t statisic
(mu_1 - mu_2 ) / SED
You can get the whole matrix of SED's mu using the !SED qualifier
predict treat !sed
Then you can use the individual SED's for each particular contrast.
Arthur
########################################
Kishore Prayaga
CSIRO Livestock Industries
Locked bag 1
Armidale
NSW 2350 Australia
Ph (02) 6776 1379 / (02) 6776 1420
Fax (02) 6776 1371
e-mail Kishore.Prayaga@csiro.au
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml