Resumen

SEGURA CORREA, José Candelario; SOLIS CALDERON, José de Jesús  y  SEGURA CORREA, Víctor Manuel. Logistic Regression Analysis For Correlated Data Using Three Procedures of the SAS Statystical System. Rev. Cient. (Maracaibo) [online]. 2006, vol.16, n.3, pp.282-287. ISSN 0798-2259.

The objective of this study was to compare the results of logistic regressions for correlated data obtained by using three procedures of the Statistical Analysis System package (SAS). Information from a cross sectional study on infectious bovine rhinotracheitis in Yucatan state, Mexico, was used; where the sampling unit was the herd and the unit of interest was the animal. Data were analyzed by using the LOGISTIC, GENMOD and NLMIXED procedures of the SAS. The logistic model used included the risk factors: herd size (£75, 76-147, 148-261, 262-480 animals) and stage of life (growth, development and production). The odd ratios (OR) and 95% confidence intervals (CI95) were obtained. Also the OR and CI95 for the risk factors were estimated using ordinary logistic regression as a reference. The ORs for herd size £75, 76-147, 148-261 y 262-480 animals, using LOGISTIC, GENMOD and NLMIXED procedures were: 0.34, 0.68, 0.62 and 1; 0.34, 0.63, 0.63 and 1; and 0.29, 0.62, 0.57 and 1, respectively; and for the animals in the stage of growth, development and production were: 0.12, 0.15 and 1; 0.15, 0.17 and 1; and 0.12, 0.14 and 1, respectively. The narrower CI95 were for the ORs from the ordinary logistic regression and the widest for the OR obtained using the NLMIXED procedure. In conclusion, the three procedures adjusted for herd effect, being the most precise those of the NLMIXED procedure.

Palabras clave : Logistic regression; marginal model; random model.