Logistic regression Flashcards
Core Idea of a logistic regression
What factors influence that something happens or not
Logistic regression assumes a ________ distribution
Binomial
What is the outcome variable in this type of multiple regression?
Binomial (could be more than 2 terms). Categorical broadly speaking.
Logistic Regression definition
Is a regression model. Specifically is a multiple regression
What is predicted in a logistic regression?
The logarithmic odds of our probability of getting a 1 or a 0
Interpretation of betas in a logistic regression
They are interpreted as the change in the log odds for every unit increase in our predictor variable
We use a logistic regression when
We have a binomial outcome variable and the outcome is also discrete
Multinomial Logistic Regression
Is a logistic regression regression where the outcome variable is multinomial; instead of yes/no you have yes/no/maybe
Iterative method
Refers to the method a logistic regression uses to determine the combination of betas that give the MAXIMUM LIKELIHOOD SOLUTION
MLE
Is the Maximum Likelihood Estimation, the computation of the best formula / line that maximize the likelihood.
Assymptotes
A function of a relationship that approaches but never reaches 1 or 0. Meaning that it tends to infinity
What does predictors do to the outcome variable
They change the log odds of an event to happen or not happen
What does the intercept mean?
Is the log odds of getting lung cancer (outcome variable) for the 0 condition
Logistic function in a logistic function
We turn log odds into odds by using this inverse function to get probabilities
Issues of a logistic regression
Overdispersion
Outliers
Separability
Backtransformation of a logistic regression
Doing the inverse of a logit link function is a logistic
Reason why we backtransform
Odds can never be negative
Logistic regression assumes violation on this assumption
Homogeneity of variace :D
R square problems
Not good because they assume homogeneity of variance
Two types of chi square
Wald
Likelihood ratio
When looking at a plot how do you know you are in log odds or probability?
If your y axis includes negative values or values greater than 0s is log odds. If it goes from 0 to 1 is probability
