Abstract

Two alternative methods for parceling questionnaire items for use in confirmatory analyses are presented. The first method requires that parcels must (a) pass a minimum standard of reliability and (b) provide indications of unidimensionality to be retained for analysis. The second method requires that parcels be equally representative of the multiple aspects of a domain. The parcels may then serve as adequate indicators for the general construct. The latter method is consistent with the rationale underlying aggregation of measures, a procedure currently recommended for improving the psychometric properties of behavioral measures of personality. The two methods for parceling and a comparison are illustrated with an empirical example.