Regression Analysis

Question 1

What is multiple regression analysis used for?

Answer 1

To describe the relationship among 2 or more variables that are interval-scaled.

Question 2

How many dependent variables do you have in MRA?

Answer 2

Single

Question 3

What is the role of least squares in MRA?

Answer 3

To establish a baseline to compare predictive ability, we examine the errors in predicting the dependent variable.

Question 4

What do the coefficients for each independent variable provide?

Answer 4

A formal basis for assessing the change in the dependent measure for each one unit change in the independent variable.

Question 5

When can you use dummy variables?

Answer 5

When the independent variable is non-metric.

Question 6

What to do when the dependent variable is non-metric?

Answer 6

Discriminant analysis or logistic regression.

Question 7

What are the (4) assumptions of MRA?

Answer 7

Linearity
Constant variance of error items
Independence of error items
Normality of the error term distribution

Question 8

What is meant by variable specification?

Answer 8

Fundamental choice between using the individual variabels or some form of composite.

Question 9

What choices do you have regarding variable selection (2)?

Answer 9

Confirmatory approach
Software-controlled approach

Question 10

What are the options available in software-controlled variable selection (4)?

Answer 10

Simultaneous regression
Combinatorial approach
All possible subsets regression
Estimation technique to pick and choose among the set of independent variables with either sequential or constrained processes

Question 11

How to interpret the result of regression and variabel importance, especially in light of multicollinearity? (2)

Answer 11

1. Examine bivariate relationship independent of the estimated model
2. Use model results to assess estimated coefficients

Question 12

What are potential influential observations (3)?

Answer 12

Outliers
Leverage points
Influential observations

Question 13

What is an advantage of multilevel models?

Answer 13

More easily handled or nested data structures

Question 14

When to use panel models?

Answer 14

Cross-sectional analyses of longitudinal or time-series data

Question 15

What are the (3) applications of MRA?

Answer 15

Analysis of causes
Forecasting impact of something
Time-series analysis

Question 16

What is the regression variate?

Answer 16

Linear combination of weighted independent variables used collectively to predict the dependent variable.

Question 17

What are the (7) steps in the analysis process of MRA?

Answer 17

Objectives
Research design
Assumptions
Estimating regression model
Assessing overall fit
Interpretation
Validation of results

Question 18

What is the minimum sample size for MRA?

Answer 18

50 and preferably 100 observations for most research situations.

Question 19

What do more degrees of freedom do? (2)

Answer 19

Improve generalizability
Addresses model parsimony and sample size concerns

Question 20

What is a potential remedy when assumption of linearity is not met?

Answer 20

Transformation of data including polynomial terms

Question 21

What does R2 stand for?

Answer 21

The measure of overall model predictive accuracy.

Question 22

What are the (3) questions to ask about statistical significance of any regression coefficient?

Answer 22

Was statistical significance established?
How does the sample size come into play?
Does it have practical significance in addition to statistical significance?

Question 23

What does the adjusted R2 do?

Answer 23

It takes into account the complexity of the model.

Question 24

What do regression coefficients describe?

Answer 24

Change in the dependent variable due to changes in independent variable, IF all other independent variables remain constant.

Question 25

When to use beta weights (standardized coefficients)?

Answer 25

When comparing relative importance among independent variables

Question 26

What are the (3) remedies in case of multicollinearity?

Answer 26

If highly correlated set of variables: use only one
Transformation into new set of independent predictors
Ridge regression and latent root regression

Question 27

How to validate results (3)?

Answer 27

Additional or split samples
Comparing regression models
Forecasting with model

Question 28

What is logistic regression?

Answer 28

Regression with a binary dependent variable.

Question 29

What does the logistic coefficient reflect (2)?

Answer 29

Direction and magnitude of the independent variables relationship

Question 30

How can the magnitude of the coefficient be determined?

Answer 30

Through evaluation of the numeric value of the coefficient.

Question 31

What are the (2) advantages of logistic regression compared to discriminant analysis and multiple regression?

Answer 31

Logistic regression is more robust
Researchers prefer straightforward statistical test of logistic regression

Question 32

What is the dependent variable in logistic regression?

Answer 32

The LOG of the odds ratio

Question 33

What are the (6) steps of a logistic regression?

Answer 33

Problem definition
Formulating the model
Estimation
Goodness of fit
Assessment
Prediction

Question 34

What do hypotheses in logistic regression look like?

Answer 34

The more… the more/less probable…

Question 35

What does the maximum likelihood do?

Answer 35

Determines the B’s so that the likelihood to receive the observed data is as high as possible.

Question 36

What test to use when the group sizes are very different?

Answer 36

The likelihood ratio test (rather than deviance).

Question 37

What are the different (3) pseudo R-square statistics?

Answer 37

McFadden
Cox and Snell
Nagelkerke