adj. describing statistical methods that have a reduced likelihood of error, particularly of incorrectly rejecting the null hypothesis, but that are less capable of detecting significant relationships among variables. In other words, conservative approaches have less power and thus tend to underestimate associations or effects.