n. the situation in which the strength of a correlation coefficient or regression equation decreases when it is applied to a new data set. Such shrinkage occurs when the initial estimate of the correlation reflects unique characteristics of the initial sample that are not replicated in subsequent samples.