|
Abstract:
|
In modeling, measuring the association between outcome variable and main covariate is the key issue. It may be biased from the effect of extraneous variable---age, for instance, which relates to the main covariate only---and with an indirect effect in the outcome variable. From point-of-data analysis, especially in several included covariates, it is not necessary to keep this extraneous variable in model. This study proposes a mean corrected method to remove this extraneous effect prior to modeling. This extraneous variable explains part of the total variation of the main covariate. The method removes this variation (among groups) through mean corrected technique. A Down syndrome study, relating to gestational age, is illustrated. The proposed method benefits from an easy approach to comparing with conditional logistic model using likelihood probability.
|
