what is the basic premise of logistic regression?
want to find the probability that P(outcome) > some threshold alpha
if multiple linear regression. what which class would have the highest probability
what does a higher or lower alpha mean for predictions?
lowering alpha will lead to more “success” predictions and fewer “failure” predictions.
why doesn’t it make sense to measure logistic regression outcomes as a linear function?
we’ll see that negative numbers are in the domain.
technically violates 0<= P(X) <= 1
how do we ensure that 0<= P(X) <= 1?
use sigmoid or logistic function
P(Z) = e^z/(1+e^z)
what is MLE used for in this context?
maximizing the regression eqn is equivilant to determining the values of beta that make the data most probable.
what is a confusion matrix?
a table that compares what the model predicts vs what is “the truth”
what would a visual of multiple logistic regression look like?
what are the advantages and disadvantages of logistic regression?
low variance but high bias
is logistic regression a flexibile model?
no its rigid bc its linear
