Abstract des Autors

In this study, generalizability theory, an extension of classic true-score reliability theory, was used to investigate the relative and absolute generalizability of the observed measurements of total body bioelectrical impedance (resistance, reactance, and phase angle) for men and women, using the bioelectrical impedance analysis (Spectrum Lightweight Instrument; RJL Systems, Detroit MI) for whole body composition. Eighty Alaskan military men (19 to 56 years old) and 83 Alaskan military women (19 to 54 years old) volunteered as subjects from Elmendorf Air Force Base, Anchorage, Alaska. The rules of RJL for testing whole body bioelectrical impedance were followed. The design was a four-faceted cross: person by day by occasion by trial. All facets were treated as random. Each subject was tested by one tester on 2 days, 2 occasions, and 3 trials. Data were analyzed independently for each impedance measurement and each sex group. The analyses showed that the persons by day interaction contributed a significant percentage of total variance on all impedance measurements for both male and female groups. The generalizability study (2 days, 2 occasions, and 3 trials) coefficient for both relative and absolute decision ranged from 0.86 to 0.96. The alternative decision study coefficient for relative decision ranged from 0.73 to 0.98, and the coefficient for absolute decision ranged from 0.72 to 0.98. Overall, these data revealed that a more precise measurement procedure can be otained by increasing the number of test days, whereas no significant enhancement may be obtained by increasing the number of trials and occasions. Verf.-Referat