STAT 101 Chapter Notes - Chapter 2: Covariate, Prenatal Care, Logistic Regression
Document Summary
Cumulative logit model (cumulative odds model, ordinal logistic model, proportiona odds model) Takes into account ordering of levels in outcome variable, actually uses ordinal info. Accommodate variables that have arbitrary but existing intervals. Modeling as a continuous when not truly continuous yields biased results-> predicted values outside range of ordinal variables. Model predicts the ln(odds of being in category j or beyond. Does not focus on specific outcome category, greater power. If an independent variable increases/decreases, then odds of being in cat 1 v all others increases/decreases proportionally. It also increases/decreases the odds of being in cat 2 and 1 combined v the remaining categories. In categories 3,2, and 1 combined v remaining categories. If holds, can express association between independent variable and outcome as. If does not hold, it might be because outcome is nominal. If does not hold, mis-specified categories, failed to find important thresholds. If does not hold, order of categories may be off.
