**Pooling**: The construction
of an error term from more than one source of variance in the response. *A priori* pooling occurs in designs
without full replication, where untestable interactions with random factors are
pooled into the residual variation. The analysis then proceeds on the
assumption that the interactions are either present (Model 1) or absent (Model
2). Planned *post hoc* pooling is
applied to mixed models by pooling a non-significant error term with its own
error term. The design is thereby influenced by the outcome of the analysis (in
terms of whether or not an error term is itself significant). More generally,
pooling can describe the process of joining together samples, for example in
calculating a main effect Mean Square by pooling across levels of a cross factor.

Doncaster, C. P. & Davey, A. J. H. (2007) *Analysis of Variance and Covariance: How to
Choose and Construct Models for the Life Sciences*. Cambridge: Cambridge
University Press.

http://www.southampton.ac.uk/~cpd/anovas/datasets/