L7 Regression with binary dependent variable

Question 1

What does an estimate of 0.5 mean when the random variable is binary

Answer 1

That there is a 50% chance of it being 1

Question 2

What is a linear regression model with a binary dependent variable called

Answer 2

A linear probability model becouse the result is a probability

Question 3

What is a probit model

Answer 3

A model that generates an S-curve between 0 and 1

Question 4

In the probit model, probability changes faster at extreme values than in the middle

Answer 4

False, it changes faster in the middle.

Question 5

Probit regression models the probability of Y=1 using a cumulative normal distribution function

Answer 5

True

Question 6

In probit regression b0 + b1x… bnx = z-value of cumulative normal distribution

Answer 6

True

Question 7

The standard errors are handled diferently in the probit model

Answer 7

False

Question 8

Is the logit function computationally faster than probit

Answer 8

yes but that matters little nowadays

Question 9

Logit and probit gives similar results

Answer 9

True

Question 10

Is R² a good measure of fit in the logit and probit model

Answer 10

No, absolutely not as the functions cannot be made linear.

Question 11

What is the accuracy rate in non linear probability regression models

Answer 11

You divide the estimates into their binary guesses and check how many percent were correct.

Question 12

When might the accuracy rate be missleading

Answer 12

When the sample data is not balanced, aka when there are way too many of positive outcomes or negative ones.

Question 13

Is there another good measure of fit for logit and probit besides the accuracy rate

Answer 13

Yes pseudo R² but it is not covered in this corse.