Testimation bias impacts the magnitude of the effect reported for a predictive factor.

Derivation: from test and estimation, indicating estimation after testing


Situation where testimation bias arises:

(1) An initial statistical test (such as determining the p value) is performed to identify possible predictive factors.

(2) Only those predictive factors showing a large effect by the statistical test are included in the final analysis.



(1) The predictors included in the final model may be reported to show a larger-than-actual effect (overestimated).

(2) Some factors that may have an effect are excluded (underestimated).


Example: Bias may be introduced if only factors with a p < 0.05 are included for analysis. This bias would not be introduced if all factors with a p < 0.2 were included instead.

To read more or access our algorithms and calculators, please log in or register.