Categorical Outcomes Flashcards
what is the equation for applying the linear model to categorical outcomes?
P = 1 /
1 + e - (b0 + b1a + b2b + b3axb)
what assumptions are involved?
linearity, sphericity, multicollinearity, incomplete information, complete separation
what is multicollinearity?
too many predicts = highly correlated with eachother
what is incomplete information?
empty cells and gaps in the data
what is complete separation?
when the outcome can be perfectly predicted, no one model best fits the data
what should you use if you have complex contingency tables with 3+ predictors?
loglinear analysis
what is Exp(b)?
the exponentiation of B, the odds ratio after a unit change in the predictor
what is Exp(b) CI?
if the CI cross 1 = no change, <1 = relationship reflects population, >1 = population is greater than relationship
how do you calculate odds?
the number of times an event occurs / the number of times an event doesn’t occur
how do you calculate an odds ratio?
odds / odds (the thing you’re looking at goes on top)
what does it mean if the odds ratio is close to 1?
no effect or no change
what is nagelkerke r-square
including the predictor, the model explains __% of the variation
what is B?
the log odds of an outcome
what is B0?
the log odds of an outcome when all predictors = 0
what is B1?
the change in log odds of an outcome, associated with a unit change in the predictor
