|
|
Dear all,
In an ordinal regression process, the logit link function uses P(lower categories) as numerator and P(higher categories) as denominator. In this case, a higher predicted log odds would actually mean higher probability of being a lower category. However, in the output of SPSS PLUM, a higher B seems to indicate a positive association between the independent variable and dependent variable, which contradicts to the definition of the link function. Moreover, the sign of intercepts seems different from other predictors. Would anyone please to suggest what's happening? Thanks so much.
Regards,
Johnson
|
Locked
