Q: Does MetaWin assess the significance of resampling tests (randomization) as one-tailed or two-tailed?
A: For both continuous and categorical models, MetaWin assesses
randomization tests as one-tailed significance tests. This is
accomplished by determining whether or not the values obtained from each
iteration are more extreme than the observed value, it its observed
direction only. A two-tailed test would compare the observed value to the
absolute value of the values from each iteration, making it a two-tailed test.
See also Corrections/Errata
Return to the MetaWin Home Page