Deck 13: Explaining Dichotomous Outcomes: Logistic Regression

Question 1

For which one of these dependent variables would you use logistic regression?

A) income in dollars
B) number of children
C) hours worked per week
D) whether or not one owns a gun

Answer 1

D

Question 2

Logistic regression uses "e", otherwise known as:

A) the exponential logarithm
B) the logistic logarithm
C) the natural logarithm
D) the regression logarithm

Answer 2

C

Question 3

If we wanted to explain work behavior, for which of the following variables would we use logistic regression?

A) hours worked
B) occupational prestige score
C) currently employed or not
D) type of industry

Answer 3

C

Question 4

According to GSS2006 data, a white woman living in a big city has a 50% of saying she is afraid to walk alone at night in her neighborhood. This percentage translated into odds is:

A) 1 to 2
B) 1 to 1
C) 2 to 1
D) 4 to 1

Question 5

If the predicted probability is 50%, what is the z-value in a logistic regression?

A) 0
B) .5
C) 1
D) 2

Question 6

In the GSS2008, respondents were asked if the aged should live with their children. 59% said this was a good idea, while 41% said it was a bad idea. Here is a logistic regression equation:
Dependent Variable: Should Aged Live with Their Children?
(1=good idea, 0 = bad idea)

$\begin{array}{l} \text { Independent Variable } & \text { Slope} \\ \text { Age of Respondent } &-.02^{* * *} \\ \text {Constant } &1.18\end{array}$ Which of the following statements is closest to the truth?

A) A ninety-year-old is three times as likely to think this is a good idea than a 20-year-old.
B) A ninety-year-old is twice as likely to think this is a good idea than a 20-year-old.
C) A ninety-year-old is half as likely to think this is a good idea than a 20-year-old.
D) A ninety-year-old is equally likely to think this is a good idea as a 20-year-old.

Question 7

Does having various types of children affect the probability of going to the zoo? Here is a logistic regression equation:
Dependent Variable: Visited Zoo in Past Year? (Yes=1, No=0)
$\begin{array}{ll}\text { Independent Variable } & \text { Slope } \\\text { \# of children less than } 6 \text { years old } & .49^{*} \\\text { \# of children } 6 \text { through } 12 \text { years old } & .37^{*} \\\text { \# of children } 13 \text { through } 17 \text { years old } & -.35^{*} \\\text { Constant } & -.02\end{array}$ Who is most likely to go to the zoo?

A) Someone with four-year-old triplets
B) Someone with nine-year-old triplets
C) Someone with fifteen-year-old triplets
D) Someone with a four-year-old, a nine-year-old, and a fifteen-year-old

Question 8

A question asked every year in the GSS is called FEAR: "Is there any area right around here - that is, within a mile - where you would be afraid to walk alone at night?" Here is a logistic regression equation using this as the dependent variable with GSS2006 data:
Dependent Variable: Afraid to Walk Alone? (No=0, Yes=1)

$\begin{array}{lc} \text { Independent Variable } & \text { Slope } \\ \text { Sex ( Male =0, Female =1) } &1.17^{* * *} \\ \text { Race (White =0 , Black =1)} &.48^{*} \\ \text { Live in City? No=0, Yes=1) } &.47^{*} \\ \text { Constant } &-1.62\end{array}$ In terms of a percentage difference, how do a white, non-city-dwelling male and a black, city-dwelling female differ?

A) They differ by 40%
B) They differ by 45%
C) They differ by 50%
D) They differ by 55%

Question 9

According to GSS2006 data, a white man living in a big city has a 25% chance of saying he is afraid to walk alone at night in her neighborhood, while a white woman living in a big city has a 50% chance of saying she is afraid of saying she is afraid to walk alone at night in her neighborhood. Which of the following is the odds ratio (men's to women's)?

A) .33
B) .5
C) 2
D) 3

Question 10

In the GSS2008, respondents were asked if the aged should live with their children. Here is a simple logistic regression equation. Race is coded as follows: White and Blacks = 0, Others (i.e. not whites or blacks) =1.
Dependent Variable: Should Aged Live with Their Children? (1=good idea, 0 = bad idea)
$\begin{array}{lc}\text { Independent Variable } & \text { Slope } \\\text { Race } & .55^{* *} \\\text { Constant } & -.10\end{array}$ What is this slope translated into an odds ratio?

A) .18
B) 1.22
C) 1.70
D) 5.00

Question 11

If the odds ratio for a gender variable (where women are coded as 1) is .33, what would the odds ratio be for the same gender variable if men had been coded as 1?

A) 2
B) 3
C) 6
D) 10

Question 12

The logistic slope for a dichotomous variable is +1.10. What is the odds ratio?

A) 1.10
B) .33
C) 0
D) 3

Question 13

What is the relationship between odds ratios and predicted probabilities?

A) They really the same thing.
B) They are usually very similar.
C) You can use only one at a time.
D) They can be quite different.

Question 14

Beyerlein, Soule, and Martin, in their research on protest, study the connections among:

A) prayers, protest, and police
B) marches, masculinity, and the military
C) sit-ins, sanctuary cities, and sentencing
D) activism, attitudes, and art

Question 15

Beyerlein, Soule, and Martin use which method in their research on protest?

A) survey research
B) secondary analysis
C) experimentation
D) content analysis

Question 16

Beyerlein, Soule, and Martin, in their research on protest, use which regression techniques?

A) logistic regression and interaction effects
B) logistic regression and nested modeling
C) interaction effects and nested modeling
D) logistic regression, interaction effects, and nested modeling

Question 17

Beyerlein, Soule, and Martin, in their research on protest, explore evidence for the concept of:

A) protest legitimacy
B) political legitimacy
C) privileged legitimacy
D) police legitimacy

Question 18

In their research on attitudes toward global warming, McCright and Dunlap use data from:

A) the General Social Survey
B) the American National Election Studies
C) the World Values Survey
D) Gallup polls

Question 19

Which combination of regression techniques to McCright and Dunlap use in their research on attitudes toward global warming?

A) logistic regression and nested modeling
B) logistic regression and interaction effects
C) nested modeling and interaction effects
D) nested modeling and standardized effects

Question 20

What aspect of global warming do McCright and Dunlap use as their dependent variable?

A) whether or not the respondent thinks the effects of global warming have begun
B) whether or not the respondent thinks humans are responsible for global warming
C) whether or not the respondent cares about global warming
D) whether or not the respondent has given money to an organization fighting global warming

Question 21

According to McCright and Dunlap, which of the following types of people is LEAST likely to have said that global warming has already begun?

A) a non-educated liberal Democrat who claims to understand global warming not at all
B) a highly-educated liberal Democrat who claims to understand global warming very well
C) a non-educated conservative Republican who claims to understand global warming not at all
D) a highly-educated conservative Republican who claims to understand global warming very well

Question 22

In SPSS, which of the following versions of R-Square does logistic regression output give you?

A) Cox and Snell
B) Nagelkerke
C) Both of these
D) Neither of these

Question 23

What does not appear in SPSS logistic regression output?

A) unstandardized slopes
B) standardized slopes
C) statistical significance
D) a measure of R-square

Question 24

In SPSS, the logistic regression output gives you a column headed "Exp(B)." What does this mean?

A) e to the power of the slope
B) e multiplied by the slope
C) the slope to the power of e
D) the expected value of the slope

Question 25

If you try to run a regular regression using a dichotomy in SPSS, SPSS will:

A) do it
B) do it, but give you a warning message
C) give you an error message
D) cease to work

Question 26

Explain why, if your dependent variable is a dichotomy, logistic regression is preferred to regular regression.

Question 27

A friend of yours uses logistic regression to try to explain variation in the number of guns people own, but SPSS keeps giving him an error when he tries to run his model. Explain why this is happening, and offer your friend a suggestion for fixing the problem.

Question 28

GSS2008 respondents were asked: "Should divorce in this country be easier or more difficult to obtain than it is now?" 48% of the respondents say it should be harder. A logistic regression equation:
Dependent Variable: Divorce Should be Harder to Obtain (Yes=1, No=0)
Based on the results, describe the two extreme possibilities (i.e. the two types of people whose probabilities would differ the most) and then calculate probabilities for both of them.

Question 29

We want to learn more about variation in attitudes towards anti-American Muslims. We notice that the 2008GSS had a new variable called LIBMSLM: "If some people in your community suggested that a book an anti-American Muslim clergyman wrote which preaches hatred of the United State should be taken out of your public library, would you favor removing this book or not?" 50% of Americans said remove it, 50% said do not remove it. Here is a logistic regression equation:
Dependent Variable: Remove Book? (Yes=0, No=1)
Based on the results, describe the two extreme possibilities (i.e. the two types of people whose probabilities would differ the most) and then calculate probabilities for both of them.

Question 30

Here is a logistic regression equation that includes an interaction effect. EDUC is measured in years, RACE is coded White=0, Blacks=1. The dependent variable is SPANKING: where 0=disagrees with spanking children, 1=in favor of spanking children (all slopes statistically significant):
SPANKING = 2.44 - 1.27(RACE) - .11(EDUC) + .16(RACE)(EDUC)
By calculating four examples (two white people: one with 10 years of education, one with 20 years of education; two black people: one with 10 years of education, one with 20 years of education), tell the story behind this model.

Question 31

The 2008GSS asked respondents if they can speak a language other than English. Using only those born in the U.S., here are logistic regression results with an interaction effect:
Dependent Variable: Can Speak Language Other than English? (No=0, 1=Yes)
Run the following examples and tell the resulting story:
a man with 10 years of education
a man with 20 years of education
a woman with 10 years of education
a woman with 20 years of education

Question 32

The GSS2008 asks people if they have seen an x-rated movie in the past year. Below is a logistic regression equation. The "Happy Marriage" variable is measured as follows: 0=Not Happy, 1=Pretty Happy, and 2=Very Happy.
Dependent Variable: Seen X-Rated Movie in Past Year? (No=0, 1=Yes)
Calculate probabilities for a 40-year-old man in a pretty happy marriage and a 40-year-old woman in a pretty happy marriage. Then, use these examples to show that e^-1.23 is the odds ratio of female odds to male odds.

Question 33

Group A has a 60% chance of owning a home, while Group B has a 20% chance of owning a home. Use these examples to show that the odds ratio of owning a home is NOT the same as the ratio of the likelihoods of owning a home.

Question 34

In the wake of 9/11, the GSS asked a new set of questions in 2006 regarding civil rights. One of them was: "Suppose the government suspected that a terrorist act was about to happen. Do you think the authorities should have the right to detain people for as long as they want without putting them on trial?" 53.5% of the respondents said yes, 46.5% said no. Here is a logistic regression model that uses two independent variables: Political Views (Extremely Liberal=0, up to Extremely Conservative=6) and Party Identification (Strong Democrat=0, up to Strong Republican=6).
Dependent Variable: OK with Indefinite Detention? No=0, Yes=1
Calculate predicted probabilities for an extremely liberal Democrat and an extremely conservative Republican.

Question 35

By describing their dependent variable, explain why Beyerlein, Soule, and Martin use logistic regression in their article on protest.