Analysis of Ordinal Categorical Data, Second Edition by Alan Agresti(auth.)

By Alan Agresti(auth.)

Statistical science’s first coordinated guide of tools for reading ordered express information, now totally revised and up-to-date, keeps to provide purposes and case experiences in fields as varied as sociology, public overall healthiness, ecology, advertising, and pharmacy. Analysis of Ordinal specific info, moment Edition presents an advent to easy descriptive and inferential tools for specific information, giving thorough insurance of latest advancements and up to date equipment. precise emphasis is put on interpretation and alertness of equipment together with an built-in comparability of the on hand techniques for examining ordinal facts. Practitioners of records in govt, (particularly pharmaceutical), and academia will wish this new edition.Content:
Chapter 1 creation (pages 1–8):
Chapter 2 Ordinal possibilities, rankings, and Odds Ratios (pages 9–43):
Chapter three Logistic Regression types utilizing Cumulative Logits (pages 44–87):
Chapter four different Ordinal Logistic Regression types (pages 88–117):
Chapter five different Ordinal Multinomial reaction types (pages 118–144):
Chapter 6 Modeling Ordinal organization constitution (pages 145–183):
Chapter 7 Non?Model?Based research of Ordinal organization (pages 184–224):
Chapter eight Matched?Pairs information with Ordered different types (pages 225–261):
Chapter nine Clustered Ordinal Responses: Marginal versions (pages 262–280):
Chapter 10 Clustered Ordinal Responses: Random results versions (pages 281–314):
Chapter eleven Bayesian Inference for Ordinal reaction information (pages 315–344):

Show description

Read Online or Download Analysis of Ordinal Categorical Data, Second Edition PDF

Similar analysis books

System Analysis and Modeling: About Models: 6th International Workshop, SAM 2010, Oslo, Norway, October 4-5, 2010, Revised Selected Papers

This e-book constitutes the throughly refereed post-proceedings of the sixth foreign Workshop on platforms research and Modeling, SAM 2010, held in collocation with types 2010 in Oslo, Norway in October 2010. The 15 revised complete papers awarded went via rounds of reviewing and development. The papers are prepared in topical sections on modularity, composition, choreography, software of SDL and UML; SDL language profiles; code new release and version alterations; verification and research; and person requisites notification.

Additional info for Analysis of Ordinal Categorical Data, Second Edition

Sample text

For example, ordinal odds ratios apply to each partial table using the cell counts {ny*} in the three-way contingency table that corresponds to "stacking up" the various partial tables. ni+\,j+\,k 1 i , ; + l,t1i + l,y·,* ' 1, . . , r - 1, j = l , . . , c - 1. 15) The response variable and explanatory variable are conditionally independent, given the control variable(s), if the population values of these odds ratios all equal 1. 1 Summary Measures of Conditional Association If the association as described by an ordinal measure is similar in each partial table, it can be useful to pool the measure values into a summary measure of Using the "permutation with general scores" analysis and choosing scores 1, 2, 3, 4, and 5.

69 being slightly out of order. 3)] using just four cells in a rectangular pattern, the values tend to be stronger for less localized odds ratios. 24. 4. When different researchers may choose different numbers of categories c for an ordinal response variable, an advantage of ordinal odds ratios that are global in Y is that their values do not usually depend much on that choice, since in each case they use the entire response scale. By contrast, each local odds ratio uses more of the response scale when c is small than when c is large.

We assume multinomial sampling over the cells of the contingency table; that is, the cell counts {n,7} have a multinomial distribution with parameters that are the cell probabilities {7r,-y·}. Then ζ has an asymptotic (large-sample) normal sampling distribution by the delta method (Bishop et al. 1975, Sec. 6). Let SE denote an estimated standard error for ζ. An approximate 100(1 — a)% confidence interval for ζ is C±z„/2(SE), where za/2 denotes the standard normal percentile with right-tail probability equal to a/2.

Download PDF sample

Rated 4.19 of 5 – based on 12 votes