Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
31 Cards in this Set
- Front
- Back
How many predictors and outcome variables are there in LR?
|
One outcome variable, 1+ predictors
|
|
What kind of measurement does the outcome variable need to be in LR?
|
Categorical (respondents allocated to discrete categories)
|
|
What is the difference between Binomial LR and Multinomial LR?
|
Binomial LR has 2 categories on the outcome variable
Multinomial LR has 3 or more categories |
|
What are the two uses of LR?
|
1. Predict which category participants will be in, based on their scores on the predictor variable(s)
2. Identify which predictor variables predict the outcome (more common) |
|
What are the three major differences between linear regression and logistic regression?
|
1. Linear outcome variable is continuous, Logistic is categorical
2. Logistic regression does not assume a normal distribution, while linear does 3. Logistic regression does not assume linear relationships between predictor variables, while linear does |
|
When would you use LR over discriminant analysis?
|
When you are interested in the predictors than the end category
|
|
While linear regression predicts scores on the outcome variable, logistic regression predicts ___________
|
Category membership
|
|
In binomial logistic regression, how do you calculate the odds of being in a category?
|
probability of category 1 / probability of category 2
|
|
What is the odds ratio? (Binomial LR)
|
An estimation of the change in odds of category membership for a 1-unit increase in the predictor
|
|
What does it mean when an odds ratio or natural log is less than 1?
|
There is a negative relationship between the predictor and outcome
|
|
What is the idea behind Log Odds?
|
To standardise the odds by making the odds of opposing events equivalent
|
|
What is the shorthand for Log Odds in the binomial regression equation?
|
b
|
|
What is the shorthand for Odds Ratio in the regression equation?
|
Exp(B)
|
|
What is the bivariate LR equation?
|
g hat = a + b1
|
|
Why is a linear regression line inappropriate for LR?
|
1. The outcome variable is categorical not continuous
2. Y is not normally distributed |
|
What is the binomial logistic regression line called?
|
Sigmoidal Curve (S-shape curve)
|
|
What are the three assumptions of LR?
|
1. Categories of outcome variable are exhaustive
2. Categories of outcome variable are mutually exclusive 3. larger samples are needed to ensure statistical accuracy |
|
How many observations should be made per predictor variable?
|
Ideally, 50 or more.
|
|
What does R-squared show?
|
How well the model fits the data
|
|
What are the two tests of R-squared?
|
1. Omnibus test of model coefficients (chi-squared)
2. Hosmer and Lemeshow Test (chi-squared) |
|
What does the omnibus test of model coefficients test?
|
How much variance is explained by all predictors
(If significant, can reject H0) |
|
What does the Hosmer and Lemeshow Test test?
|
The agreement between observed and predicted outcomes
(If significant, model does not provide good fit) |
|
What is the null hypothesis for log odds and odds ratio?
|
log odds = 0
odds ratio = 1 |
|
What is the test for log odds?
|
Wald test (chi-squared)
|
|
What is the test for odds ratio?
|
Confidence Intervals for Exp(B)
|
|
Why do you need to show proportion of correct classifications?
|
It provides a general picture of the accuracy of your regression equation
|
|
What are the three methods of LR?
|
1. Direct or Enter Method
2. Sequential LR 3. Stepwise LR |
|
How are the predictors entered into the LR equation in the DIRECT or ENTER METHOD?
|
Predictors are entered in simultaneously
|
|
How are the predictors entered into the LR equation in SEQUENTIAL LR?
|
Predictors are entered in different blocks (like HMR)
|
|
When would you use STEPWISE LR?
|
When you are exploring the data in order to generate hypotheses
|
|
What are the two types of STEPWISE LR?
|
Forward methods and backward methods
|