Lecture 12- Categorical Outcomes Flashcards
What’s the equation for the probability of the outcome occurring
(P(Ŷ) / 1-P(Ŷ)) = b0+b1X+e
What is the difference between the logistic regression equation and the linear model
In the logistic regression equation we predict the log odds of the outcome
What is b1 in the logistic regression equation
- The change in the log odds of the outcome associated with a unit change in the predictor
- Easier to interpret exp(b1), the odds ratio associated with a unit change in the predictor
What’s the odds ratio exp(b)
Exponent of b
Odds after a unit change in the predictor / original odds
If >1: Predictor high
Probability of outcome occurring high
If <1: Predictor high
Probability of outcome occurring low
When building the model what does forced entry mean
All variables entered simultaneously
When building the model what does hierarchical mean
Variables entered in blocks (blocks should be based on past research or theory being tested)
When building the model what does stepwise mean
- Variables entered on the basis of statistical criteria.
- Should be used only for exploratory analysis
What can go wrong with logistic regression model
- Linearity
- Spherical residuals
- Independent errors
- Multicollinearity
- Incomplete information
- Complete seperation
What’s the problem with empty cells
- Inflates standard errors
- Problem quickly escalates with continuous predictors
Why would you predict the probability of outcome occurring
Because a linear model can’t be fit (aka categorical data)
The exponent converts the logarithm
Back to it’s original value
What is b0 in the logarithm equation
- Log odds of outcome when the predictors are 0
- Easier to interpret exp(b0), the odds of the outcome when predictor is 0
