Ordinal outcome variables

Jimmie Leppink

doi:10.1007/978-3-030-21241-4_7

Ordinal outcome variables

Jimmie Leppink

The Hull York Medical School

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract

A commonly undervalued and mistreated type of outcome variable is the ordinal one. Two common types of mistreatment are treating ordinal variables as interval/ratio level outcome variables (frequently in linear models) and, in other cases, as multicategory nominal outcome variables. Multicategory nominal outcome variables are covered in Chap. 6 and quantitative outcome variables in Chap. 8 of this book. What these two types of mistreatment have in common is that they more often than not may result in outcomes that do not make sense. Where we treat ordinal variables as if they were (at least) interval, we may see linear relations where they do not make sense. Where we treat ordinal variables as if they were nominal, we treat all categories as exchangeable and lose the information with regard to the order of categories and, to a large extent, the meaning of outcomes with it. Although many of the concepts discussed in the context of dichotomous (Chap. 5) and multicategory nominal outcome variables also have their use, we need to take an additional step to respect the ordinality information when dealing with ordinal outcome variables. Differences between approaches to ordinal data encountered in the literature are discussed first in terms of animal comparisons (mice, hedgehogs, cats, bears, and elephants) and then in the form two experiments each of which indicates a different type of treatment effect.

Original language	English
Title of host publication	Statistical methods for experimental research in education and psychology
Place of Publication	Switzerland
Publisher	Springer
Chapter	7
Pages	103-116
Number of pages	14
ISBN (Electronic)	9783030212414
ISBN (Print)	978-3-030-21240-7
DOIs	https://doi.org/10.1007/978-3-030-21241-4_7
Publication status	Published - 2019

Publication series

Name	Springer Texts in Education
Publisher	Springer
ISSN (Print)	2366-7672
ISSN (Electronic)	2366-7680

Access to Document

10.1007/978-3-030-21241-4_7

Cite this

@inbook{06e1b9046dcc48ff80f2d20b755f6d14,

title = "Ordinal outcome variables",

abstract = "A commonly undervalued and mistreated type of outcome variable is the ordinal one. Two common types of mistreatment are treating ordinal variables as interval/ratio level outcome variables (frequently in linear models) and, in other cases, as multicategory nominal outcome variables. Multicategory nominal outcome variables are covered in Chap. 6 and quantitative outcome variables in Chap. 8 of this book. What these two types of mistreatment have in common is that they more often than not may result in outcomes that do not make sense. Where we treat ordinal variables as if they were (at least) interval, we may see linear relations where they do not make sense. Where we treat ordinal variables as if they were nominal, we treat all categories as exchangeable and lose the information with regard to the order of categories and, to a large extent, the meaning of outcomes with it. Although many of the concepts discussed in the context of dichotomous (Chap. 5) and multicategory nominal outcome variables also have their use, we need to take an additional step to respect the ordinality information when dealing with ordinal outcome variables. Differences between approaches to ordinal data encountered in the literature are discussed first in terms of animal comparisons (mice, hedgehogs, cats, bears, and elephants) and then in the form two experiments each of which indicates a different type of treatment effect.",

author = "Jimmie Leppink",

year = "2019",

doi = "10.1007/978-3-030-21241-4_7",

language = "English",

isbn = "978-3-030-21240-7",

series = "Springer Texts in Education",

publisher = "Springer",

pages = "103--116",

booktitle = "Statistical methods for experimental research in education and psychology",

address = "Germany",

}

TY - CHAP

T1 - Ordinal outcome variables

AU - Leppink, Jimmie

PY - 2019

Y1 - 2019

N2 - A commonly undervalued and mistreated type of outcome variable is the ordinal one. Two common types of mistreatment are treating ordinal variables as interval/ratio level outcome variables (frequently in linear models) and, in other cases, as multicategory nominal outcome variables. Multicategory nominal outcome variables are covered in Chap. 6 and quantitative outcome variables in Chap. 8 of this book. What these two types of mistreatment have in common is that they more often than not may result in outcomes that do not make sense. Where we treat ordinal variables as if they were (at least) interval, we may see linear relations where they do not make sense. Where we treat ordinal variables as if they were nominal, we treat all categories as exchangeable and lose the information with regard to the order of categories and, to a large extent, the meaning of outcomes with it. Although many of the concepts discussed in the context of dichotomous (Chap. 5) and multicategory nominal outcome variables also have their use, we need to take an additional step to respect the ordinality information when dealing with ordinal outcome variables. Differences between approaches to ordinal data encountered in the literature are discussed first in terms of animal comparisons (mice, hedgehogs, cats, bears, and elephants) and then in the form two experiments each of which indicates a different type of treatment effect.

AB - A commonly undervalued and mistreated type of outcome variable is the ordinal one. Two common types of mistreatment are treating ordinal variables as interval/ratio level outcome variables (frequently in linear models) and, in other cases, as multicategory nominal outcome variables. Multicategory nominal outcome variables are covered in Chap. 6 and quantitative outcome variables in Chap. 8 of this book. What these two types of mistreatment have in common is that they more often than not may result in outcomes that do not make sense. Where we treat ordinal variables as if they were (at least) interval, we may see linear relations where they do not make sense. Where we treat ordinal variables as if they were nominal, we treat all categories as exchangeable and lose the information with regard to the order of categories and, to a large extent, the meaning of outcomes with it. Although many of the concepts discussed in the context of dichotomous (Chap. 5) and multicategory nominal outcome variables also have their use, we need to take an additional step to respect the ordinality information when dealing with ordinal outcome variables. Differences between approaches to ordinal data encountered in the literature are discussed first in terms of animal comparisons (mice, hedgehogs, cats, bears, and elephants) and then in the form two experiments each of which indicates a different type of treatment effect.

U2 - 10.1007/978-3-030-21241-4_7

DO - 10.1007/978-3-030-21241-4_7

M3 - Chapter

SN - 978-3-030-21240-7

T3 - Springer Texts in Education

SP - 103

EP - 116

BT - Statistical methods for experimental research in education and psychology

PB - Springer

CY - Switzerland

ER -