CBR~PAEEATIVE VALIDITY T'S. UTILITY AS COKSEDERATIONS IN PERSONALITY SCALE CONSTRZTCTION
Herbert W. Eber · 1999
The best'' strategy for personality scale constl.uctio~: may be a utilitarian concept rather than a ps>~c.hornek~ic one. Some specific suggesaloza are giver., In his MBR monograph, Goldberg f 1972) dealt extex~si~~ely with the problem of differentia1 validities of scales that had been extracted from a single item pool by different rnethocls. He found little srtperiority for any particular types of scales, but noted that relatively broact band scales survi~ed cross validation better than did those that were more narrowly co1.1ceived. If I read his data COFFQC~I~, and If I may be permitted to benefit from hindsight, I see a rationality in his findirrgs and those of a number of subsequent ~nrestigators. A brief discussion of them may be helpfrrl. Assume that a group of items spans, in relation tcr a certain set of criteria, a definecf space. Assume that the items are assembled into scales by any neth hod other than a determined effort to suppress predictive variance by mutual cancellation. It. seems reasonable to believe that viaual!y any such aggregates of items would perform about equally well in predicting the criteria, st least in the initial sample. Assume frrrther that as is inevitable, some capitaIization upon chance has occurred in the development of dl the predictor equations. Cross validation should then find the best results from those scales which capitalize leasjt upon such chance occurrence, obviously the broader scales, the Ionger ones, the less rigidly defined ones. Assume further that the size of the samples upon which this process is carried out is vastly increased, so that the narrower and more specific scales now capitalize less upon chance. It \\-ouid be reasonable to assume that, in the asymptotic case of infinitely large sttnrples, the difference in the scales would disappear, These might be slight advantage in the more specific scales, since whatever ndclitionai validity they hare achieved wouId be preserved on cross vafiEditor's note. The stimulus for this article was the author's redew of the preceding itrticIe.