Elementary Signal Detection Theory

by Thomas D. Wickens

Overview

Signal detection theory, as developed in electrical engineering and based on statistical decision theory, was first applied to human sensory discrimination about 40 years ago. The theory's intent was to explain how humans discriminate and how we might use reliable measures to quantify this ability. An interesting finding of this work is that decisions are involved even in the simplest of discrimination tasks--say, determining whether or not a sound has been heard (a yes-no decision). Detection theory has been applied to a host of varied problems (for example, measuring the accuracy of diagnostic systems, survey research, reliability of lie detection tests) and extends far beyond the detection of signals. This book is a primer on signal detection theory, useful for both undergraduates and graduate students.

Product Details

ISBN-13:	9780190282417
Publisher:	Oxford University Press
Publication date:	10/11/2001
Sold by:	Barnes & Noble
Format:	eBook
File size:	8 MB

About the Author

Thomas D. Wickens is Professor of Psychology at the University of California, Los Angeles.

Table of Contents

1	The signal-detection model	3
1.1	Some examples	3
1.2	Hits and false alarms	6
1.3	The statistical decision representation	9
	Reference notes	15
	Exercises	16
2	The equal-variance Gaussian model	17
2.1	The Gaussian detection model	17
2.2	The equal-variance model	20
2.3	Estimating d' and [lambda]	22
2.4	Measuring bias	26
2.5	Ideal observers and optimal performance	32
	Reference notes	36
	Exercises	37
3	Operating characteristics and the Gaussian model	39
3.1	The operating characteristic	39
3.2	Isocriterion and isobias contours	42
3.3	The equal-variance Gaussian operating characteristic	45
3.4	The unequal-variance Gaussian model	48
3.5	Fitting an empirical operating characteristic	52
3.6	Computer programs	56
	Reference notes	58
	Exercises	58
4	Measures of detection performance	60
4.1	The distance between distributions	61
4.2	Distances to the isosensitivity line	64
4.3	The area under the operating characteristic	66
4.4	Recommendations	72
4.5	Measures of bias	74
4.6	Aggregation of detection statistics	78
	Reference notes	81
	Exercises	81
5	Confidence ratings	83
5.1	The rating experiment	83
5.2	The detection model for rating experiments	85
5.3	Fitting the rating model	88
	Exercises	91
6	Forced-choice procedures	93
6.1	The forced-choice experiment	93
6.2	The two-alternative forced-choice model	96
6.3	Position bias	98
6.4	Forced-choice and yes/no detection tasks	104
6.5	The K-alternative forced-choice procedure	106
	Exercises	111
7	Discrimination and identification	113
7.1	The two-alternative discrimination task	114
7.2	The relationship between detection and discrimination	118
7.3	Identification of several stimuli	124
	Reference notes	129
	Exercises	129
8	Finite-state models	131
8.1	The high-threshold model	131
8.2	The high-threshold operating characteristic	137
8.3	Other finite-state representations	140
8.4	Rating-scale data	143
	Reference notes	148
	Exercises	148
9	Likelihoods and likelihood ratios	150
9.1	Likelihood-ratio tests	151
9.2	The Bayesian observer	157
9.3	Likelihoods and signal-detection theory	160
9.4	Non-Gaussian distributions	165
	Reference notes	168
	Exercises	169
10	Multidimensional stimuli	172
10.1	Bivariate signal detection	172
10.2	Likelihood ratios	176
10.3	Compound signals	179
10.4	Signals with correlated components	184
10.5	Uncertainty effects	188
	Reference notes	192
	Exercises	193
11	Statistical treatment	195
11.1	Variability in signal-detection studies	195
11.2	Fundamental sampling distributions	198
11.3	Simple detection statistics	201
11.4	Confidence intervals and hypothesis tests	206
11.5	Goodness-of-fit tests	212
11.6	Comparison of hierarchical models	217
11.7	Interobserver variability	221
	Reference notes	223
	Exercises	224
Appendix A	Summary of probability theory	226
A.1	Basic definitions	226
A.2	Random variables	229
A.3	Some specific distributions	235
	References	253
	Index	257

Editorial Reviews

"This book contains the theoretical explications of the ways observers detect weak, uncertain, or ambiguous signals. It explains the math underlying the theory, and outlines its uses in measuring an observer's sensitivity. The book is intended to serve as an introductory text for undergraduate or graduate courses in sensation and perception, psychophysics, cognition, and quantitative methods; it may also be used as a reference for researchers. Wickens teaches at the University of California, Los Angeles."—SciTech Book News

From the Publisher