Check this box to shrink the
variance
components estimates using a prior
distribution
estimated from the data.
The prior distribution is determined empirically across all rows by fitting an inverted
gamma
distribution to row-by-row estimates. When such a prior is appropriate, the variance component estimates are more stable, and there is a fractional increase in degrees of freedom for
t
- and
F
-tests.
Tip
: This option can result in improved
power
and
sensitivity
of the analysis, especially for small data sets.