Fitting a Logistic Regression Model

Question 1

What method does PROC LOGISTIC use for estimating the unknown parameters in a logistic regression model?

Answer 1

PROC LOGISTIC uses the maximum likelihood method for estimating the unknown parameters in a logistic regression model.

Question 2

TRUE or FALSE: Higher percentages of concordant pairs and lower percentages of discordant and tied pairs indicate a more desirable logistic regression model

Answer 2

TRUE

Question 3

PROC LOGISTIC is appropriate for what type of response variable?

Answer 3

binary, ordinal, or nominal

Question 4

What does the maxpoints=none option do when using PROC LOGISTIC?

Answer 4

The maxpoints=none option requests that no observations be displayed in the effect plots

Question 5

What PROC LOGISTIC option displays confidence intervals on the plots?

Answer 5

CLBAND

Question 6

What do goodness of fit measures allow you to do?

Answer 6

Compare one model to another (The smaller the values of the goodness-of-fit statistics, the better fit of the model to the data.)

Question 7

Which PROC LOGISTIC table shows which input variables are statistically significant, controlling for all of the other input variables in the model. ?

Answer 7

The Type 3 Analysis of Effects table

Question 8

What can be used to give an approximate ranking of the relative importance of the input variables on the fitted logistic regression model?

Answer 8

The absolute value of the standardized estimates

Question 9

What does the estimate in maximum likelihood estimate table measure ?

Answer 9

the rate of change in the logit

Question 10

The Association of Predicted Probabilities and Observed Responses table provides what kind of information?

Answer 10

several measures that assess the predictive ability of the model

Question 11

How should you interpret the four rank correlation indexes: Somers’ D, Gamma, Tau-a, and the c statistic?

Answer 11

a model with higher values for these indexes– the maximum value is one– has better predictive ability then a model with lower values for these indexes.

Question 12

What is posterior probability?

Answer 12

The probability that y equals 1 given the inputs. The term posterior means that the probability is calculated after you provide the input information to the model.

Question 13

Logistic regression is a special form of what kind of model?

Answer 13

A generalized linear model

Question 14

What is an odds ratio?

Answer 14

The odds ratio is the ratio of two odds.

The odds ratio compares the odds of the event in one group to the odds of the event in another group.

Question 15

What does the logit transformation do?

Answer 15

The logit transformation takes the log of the odds.

Question 16

What does the parameter estimate measure?

Answer 16

The parameter estimates measure the rate of change in the logit

Question 17

What PROC LOGISTIC statement enables you to obtain an odds ratio estimate for a specified change in a predictor variable?

Answer 17

the units statement

Question 18

What PROC LOGISTIC statement enables you to obtain an odds ratio estimate for a specified change in a predictor variable?

Answer 18

the UNITS statement

Question 19

How does the offset correct for oversampling?

Answer 19

The intercept is reduced by the offset value

Question 20

What is the process of detecting nonlinear relationships between the target variable and the input variables referred to as?

Answer 20

Variable screening

Question 21

What do we hope to accomplish by variable screening/

Answer 21

We hope that variable screening will further reduce the number of inputs by identifying those that are clearly irrelevant.

Question 22

What statistics should you request on a PROC CORR statement as part of variable screening?

Answer 22

Spearman correlation statistic (SPEARMAN option) Hoeffding’s D statistic (HOEFFDING option)

Question 23

A high Spearman rank and low Hoeffding D rank suggests what type of relationship with the target?

Answer 23

A non-linear relationship (non-monotonic) When the Spearman rank is high, the association is monotonic, regardless of whether the Hoeffding value is high or low.

Question 24

A low Spearman rank and high Hoeffding D rank suggests what type of relationship with the target?

Answer 24

A linear relationship (monotonic)

Question 25

What is a monotonic relationship?

Answer 25

one variable consistently increases or consistently decreases with respect to the other variable

Question 26

TRUE or FALSE: A monotonic relationship can be linear or curvilinear.

Answer 26

TRUE

Question 27

TRUE or FALSE: A linear model can adequately model a monotonic relationship, whether it is linear or curvilinear.

Answer 27

TRUE

Question 28

Why is the Spearman correlation statistic used in place of the Pearson correlation statistic during variable screening?

Answer 28

Spearman is a better choice for this technique because it is less sensitive to nonlinearities and outliers.

Question 29

What type of variable is input variables are used in variable screening?

Answer 29

The variables must be ordinal or you must be able to treat them as ordinal (for example, continuous or binary variables).

Question 30

A low Spearman rank and low Hoeffding D rank suggests what type of relationship with the target?

Answer 30

A weak association indicating an irrelevant input