Análisis de la confiabilidad de FACES III (versión en español)

OBJECTIVE: To determine through five methods (Cronbach's alpha, split, Guttman method, parallel and strict parallel) the confidence coefficient of FACES III, instrument validated on spanish version, and to evaluate it's consistency with these five statistical methods. DESIGN: Descriptive, cross-sectional. SETTING: Tlalpan Area, south of Federal District, Mexico, divided into geo-statistical zones. PATIENTS AND OTHER PARTICIPANTS: A randomised sample of 270 dwellings with proportional coverage, based on 17,895 ordinary dwellings of the area. INTERVENTION: Domicile survey using FACES III (Family Adaptability and Cohesion Evaluation Scales); April-May 1995. MEASUREMENTS AND MAIN RESULTS: The confidence coefficient of FACES III spanish version was calculated using no standardised Cronbach's Alpha = 0.69; split I = 0.73 and split II = 0.66; Guttman method = 0.75; parallel = 0.69 and strict parallel = 0.53. CONCLUSIONS: Cronbach's coefficient demonstrated more advantages than the other statistical methods in relation with the ordinal measurement scale of FACES III. We suggest to take into consideration three important aspects for the correct confidence analysis of this kind of instruments: Variance analysis depending on the scale of items (F test, Friedman or Cochran) Inter-items interactions analysis (nonadditivity) and the use of the balance value as purge element of error' source. Analysis of Tukey estimate, coefficient of concordance (W) and Hotelling's T squared. It's necessary the evaluation and analysis of this aspects before the report of confidence's coefficient values whose can have hidden skew.