Final-lecture 16 Flashcards
What do logistic and linear regression try to do?
-predict the DV based on one or more independent variables
What does linear regression model?
-interval-ratio dependent variable
What does logistic regression model?
- categorical variables
- creates a linear regression using dummy variables
What are the two types of logistic regression?
- binary logistic regression
- multinomial (polychotomous) logistic regression
What is binary logistic regression?
- when dependent categorical variables have two possible outcomes
- dependent variable becomes a dichotomous dummy variable coded as 0 and 1
What is a multinomial logistic regression?
- categorical variable with more than two categories
- not within our class though
Why use logistic regression?
- many important research topics in sociology for which the dependent variable has been coded as a dichotomous nominal variable
- did you vote (yes or no)
What happens when the dependent variable is categorical?
-the assumption of linearity between the DV and IV that is assumed in OLS regression is violated
Can we use mean in logistic regression?
-no
How are logistic regression results interpreted?
-using OR
What is the common logarithm?
-base 10
Why do we have to use logistic regression and not linear for a dichotomous dependent variable?
- because logistic regression fits the line in between 0 and 1
- since probability of an event ranges from 0 to 1
What type of curve does the logistic regression model use?
- s-shaped
- sigmoid or logistic curve
Does b in logistic regression have a clear cut interpretation?
- no
- since it is a curvilinear relationship between probability and X
How do we linearize the logistic regression model?
-transforming the dependent variable from probability to logit
What is the logit in the formula called?
- a link function since it provides a linear transformation
- logistic regression is considered a generalized linear model
How do OLS and logistic regression relate in interpretation?
- relationship is linear for both
- thus, a and b coefficients are interpreted the same way
- but Y-axis is logits
What is b in logistic regression?
-the amount of change in the predicted log odds for a one unit change in the independent variable
What does a tell us in a logistic regression?
-the value of the log odds when X is zero
What do we do to b to make it more sensible?
-anti log it so that it is the effects on the odds scale
