Hello,
This is not an SPlus specific question but, I value the opinions of
the subscribers here and hope I do not offend by getting off topic.
Say, I have an experimental design with: response Y, treatment T and
blocks B.
I have thus far been writing this model in the article text as:
Y_ij = mu + T_i + B_j + epsilon_ij, i = 1:6, j = 1:3,
where the TB interaction is implicit in the epsilon error term.
However, the senior author insists on a form such as:
Y_ij = mu + T_i + B_j + TB_ij + epsilon, i = 1:6, j = 1:3
Which form is correct, or *more* correct if you will?
I fear that explicitly stating the TB interaction is duplicating this
term.
The other opinion is that the epsilon (no subscripts) is the residual
error
not captured by the design and the TB_ij interaction term is the
experimental
error *manipulated* by the design.
Extending this to a split-plot design, does one declare two different
epsilons
(with different subscripts) in the first form?
Or one epsilon (no subscripts) in the second form?
Many thanks in advance, DaveT.
**********************************************************
Silviculture Data Analyst
Ontario Forest Research Institute
Ontario Ministry of Natural Resources
Sault Ste. Marie, Ontario, Canada
david.thompson@mnr.gov.on.ca
**********************************************************
|