Log-linear models / David Knoke, Peter J. Burke.

Format:

Book

Author/Creator:

Knoke, David.

Burke, Peter J. (Peter James), 1939- author.

Series:

Sage university paper

A Sage university paper.

Quantitative applications in the social sciences ; 20

Language:

English

Subjects (All):

Social sciences--Mathematical models.

Social sciences.

Political sociology--Mathematical models.

Political sociology.

Log-linear models.

Physical Description:

79 pages : illustrations ; 22 cm.

Place of Publication:

Beverly Hills, Calif. : Sage Publications, [1980]

Summary:

Introduces methods for quantitative assessment of relationships among categoric variables in multivariable crosstabulations. Procedures to estimate and interpret effect parameters for hierarchical models are described for both the general loglinear model and its logit version.

Contents:

1. Relationships in Crosstabulations 8

2. The Log-Linear Model 11

A. Specifying Models 11

Saturated Models 12

Nonsaturated Models 17

B. Fitting Marginals 19

Generating Expected Frequencies 22

C. Analyzing Odds 24

3. Testing for Fit 30

A. How To Evaluate Models Fitted to Data 30

B. Comparisons of Different Models of the Same Data 31

Independence Hypothesis 32

Equal Marginal Distributions Hypothesis 33

C. More Complex Models: Polytomous Variables 33

D. More Complex Hypotheses 37

E. An Analog to Multiple R[superscript 2] for Large Samples 40

4. Applications to Substantive Problems 42

A. Causal Models for Log-Linear Models 42

B. Analyzing Change Over Time 47

Comparative Cross-Sections 47

Two-Wave Panels 48

Markov Chain Models 54

Age, Period, and Cohort Models 57

5. Special Techniques with Log-Linear Models 63

A. What To Do About Zero Cells 63

B. Fixing Start Values 65

C. Analyzing Ordered Data 67

D. Collapsing Polytomous Variables 70

E. Nonhierarchical Models 72.

Notes:

Bibliography: pages 77-79.

ISBN:

080391492X

OCLC:

6421394