To modify an estimate by moving it towards a benchmark. For example, one could shrink <a href="search.pl?p=parameters">parameters</a> of a model based on an individual product toward those of a model built from a general class of products. In general, shrinking reduces the size of effects (e.g., the trend or the difference from a typical case). Shrinking is useful where <a href="search.pl?p=uncertainty">uncertainty</a> exists. (See <a href="search.pl?p=mitigation">mitigation</a>.) <o:p></o:p></P>