Principal component analysis of d-prime values from sensory discrimination tests using binary paired comparisons

Christine Borgen Linander, Rune Haubo Bojesen Christensen, Graham Cleaver, Per Bruun Brockhoff

Research output: Contribution to journalJournal articleResearchpeer-review

4 Downloads (Pure)

Abstract

When considering sensory discrimination studies, multiple d-prime values are often obtained from several sensory attributes. In this paper, we introduce principal component analysis as a way of gaining information about d-prime values across sensory attributes. Specifically, we propose estimating d-prime values using a Thurstonian mixed model for binary paired comparison data and then using these estimates in a principal component analysis. Binary paired comparisons are a sensitive way to test products with only subtle differences. When analyzing data with a Thurstonian mixed model, product-specific as well as assessor-specific d-prime values are obtained. Principal component analysis of these values results in information about products and assessors across multiple sensory attributes illustrated by product and attribute maps. Furthermore, the analysis captures individual differences. Thus, by using d-prime values from a multi-attribute 2-AFC study in principal component analysis insights that are typically obtained considering quantitative descriptive analysis are obtained.
Original languageEnglish
Article number103864
JournalFood Quality and Preference
Volume81
Number of pages9
ISSN0950-3293
DOIs
Publication statusPublished - 2020

Keywords

  • d-prime values
  • Discrimination testing
  • Assessor information
  • Multi-product setting
  • Principal Component Analysis

Cite this

@article{783f9536fa924228ae0f1fdd9325643e,
title = "Principal component analysis of d-prime values from sensory discrimination tests using binary paired comparisons",
abstract = "When considering sensory discrimination studies, multiple d-prime values are often obtained from several sensory attributes. In this paper, we introduce principal component analysis as a way of gaining information about d-prime values across sensory attributes. Specifically, we propose estimating d-prime values using a Thurstonian mixed model for binary paired comparison data and then using these estimates in a principal component analysis. Binary paired comparisons are a sensitive way to test products with only subtle differences. When analyzing data with a Thurstonian mixed model, product-specific as well as assessor-specific d-prime values are obtained. Principal component analysis of these values results in information about products and assessors across multiple sensory attributes illustrated by product and attribute maps. Furthermore, the analysis captures individual differences. Thus, by using d-prime values from a multi-attribute 2-AFC study in principal component analysis insights that are typically obtained considering quantitative descriptive analysis are obtained.",
keywords = "d-prime values, Discrimination testing, Assessor information, Multi-product setting, Principal Component Analysis",
author = "Linander, {Christine Borgen} and Christensen, {Rune Haubo Bojesen} and Graham Cleaver and Brockhoff, {Per Bruun}",
year = "2020",
doi = "10.1016/j.foodqual.2019.103864",
language = "English",
volume = "81",
journal = "Food Quality and Preference",
issn = "0950-3293",
publisher = "Pergamon Press",

}

Principal component analysis of d-prime values from sensory discrimination tests using binary paired comparisons. / Linander, Christine Borgen; Christensen, Rune Haubo Bojesen; Cleaver, Graham; Brockhoff, Per Bruun.

In: Food Quality and Preference, Vol. 81, 103864, 2020.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - Principal component analysis of d-prime values from sensory discrimination tests using binary paired comparisons

AU - Linander, Christine Borgen

AU - Christensen, Rune Haubo Bojesen

AU - Cleaver, Graham

AU - Brockhoff, Per Bruun

PY - 2020

Y1 - 2020

N2 - When considering sensory discrimination studies, multiple d-prime values are often obtained from several sensory attributes. In this paper, we introduce principal component analysis as a way of gaining information about d-prime values across sensory attributes. Specifically, we propose estimating d-prime values using a Thurstonian mixed model for binary paired comparison data and then using these estimates in a principal component analysis. Binary paired comparisons are a sensitive way to test products with only subtle differences. When analyzing data with a Thurstonian mixed model, product-specific as well as assessor-specific d-prime values are obtained. Principal component analysis of these values results in information about products and assessors across multiple sensory attributes illustrated by product and attribute maps. Furthermore, the analysis captures individual differences. Thus, by using d-prime values from a multi-attribute 2-AFC study in principal component analysis insights that are typically obtained considering quantitative descriptive analysis are obtained.

AB - When considering sensory discrimination studies, multiple d-prime values are often obtained from several sensory attributes. In this paper, we introduce principal component analysis as a way of gaining information about d-prime values across sensory attributes. Specifically, we propose estimating d-prime values using a Thurstonian mixed model for binary paired comparison data and then using these estimates in a principal component analysis. Binary paired comparisons are a sensitive way to test products with only subtle differences. When analyzing data with a Thurstonian mixed model, product-specific as well as assessor-specific d-prime values are obtained. Principal component analysis of these values results in information about products and assessors across multiple sensory attributes illustrated by product and attribute maps. Furthermore, the analysis captures individual differences. Thus, by using d-prime values from a multi-attribute 2-AFC study in principal component analysis insights that are typically obtained considering quantitative descriptive analysis are obtained.

KW - d-prime values

KW - Discrimination testing

KW - Assessor information

KW - Multi-product setting

KW - Principal Component Analysis

U2 - 10.1016/j.foodqual.2019.103864

DO - 10.1016/j.foodqual.2019.103864

M3 - Journal article

VL - 81

JO - Food Quality and Preference

JF - Food Quality and Preference

SN - 0950-3293

M1 - 103864

ER -