Premium
Distributional Shapes and Validity Transport: A comparison of lower bounds
Author(s) -
Kisamore Jennifer L.
Publication year - 2008
Publication title -
international journal of selection and assessment
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.812
H-Index - 61
eISSN - 1468-2389
pISSN - 0965-075X
DOI - 10.1111/j.1468-2389.2008.00406.x
Subject(s) - credibility , variance (accounting) , external validity , econometrics , statistics , compromise , upper and lower bounds , limiting , distribution (mathematics) , normal distribution , measure (data warehouse) , psychology , mathematics , statistical physics , computer science , mathematical analysis , data mining , economics , physics , mechanical engineering , social science , accounting , sociology , political science , law , engineering
Organizations use validity transport to assess whether adoption of a particular selection measure may be of value. Current validity transport methodology assumes a normal distribution of validity parameters. Research over the past two decades has questioned this assumption, investigating the types of decision errors likely with various non‐normal parameter distributions. The current paper demonstrates that the variance of the parameter distribution and choice of lower bound has a greater impact on the true lower bound of the parameter distribution than does its shape. The current practice of using 80% credibility intervals provides a reasonable compromise in terms of invariance to parameter distribution shape while also limiting the probability of erroneously concluding validity transport is reasonable in a specific case.