Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity

Greg Atkinson, Alan M. Batterham, Helen Jones, Chloe E. Taylor, Chris K. Willie, Yu Chieh Tzeng

Research output: Contribution to journalArticleResearchpeer-review

12 Citations (Scopus)

Abstract

An individual's baroreflex sensitivity is typically described by the relationship between serial manipulations in systolic blood pressure and the changes in pulse interval. Although this experimental approach is essentially within-subjects in nature, least squares regression (LSR) analysis is typically employed by researchers to derive sensitivity slopes (gains) for individual subjects. These individual gains are then pooled as summary measures for various samples or experimental conditions. We highlight that the underlying assumption for LSR of case independence is violated with such an approach, resulting in possible estimation biases and compromised statistical power. Using a typical data set, we introduce more appropriate analyses based on the linear mixed model, which takes into account the correlated nature of the data at the individual subject level. We encourage researchers to consider the linear mixed model approach because it is more efficient, in that the whole data set is analysed in one step, is associated with less bias and results in greater statistical power compared with conventional analyses of baroreflex sensitivity for samples of subjects.

Original languageEnglish
Pages (from-to)80-82
Number of pages3
JournalClinical Physiology and Functional Imaging
Volume31
Issue number1
DOIs
Publication statusPublished - 1 Jan 2011

Fingerprint

Baroreflex
Statistical Models
Least-Squares Analysis
Linear Models
Research Personnel
Blood Pressure
Regression Analysis
Datasets

Cite this

Atkinson, Greg ; Batterham, Alan M. ; Jones, Helen ; Taylor, Chloe E. ; Willie, Chris K. ; Tzeng, Yu Chieh. / Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity. In: Clinical Physiology and Functional Imaging. 2011 ; Vol. 31, No. 1. pp. 80-82.
@article{82f6abaa287a4422b57a209a225d9c84,
title = "Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity",
abstract = "An individual's baroreflex sensitivity is typically described by the relationship between serial manipulations in systolic blood pressure and the changes in pulse interval. Although this experimental approach is essentially within-subjects in nature, least squares regression (LSR) analysis is typically employed by researchers to derive sensitivity slopes (gains) for individual subjects. These individual gains are then pooled as summary measures for various samples or experimental conditions. We highlight that the underlying assumption for LSR of case independence is violated with such an approach, resulting in possible estimation biases and compromised statistical power. Using a typical data set, we introduce more appropriate analyses based on the linear mixed model, which takes into account the correlated nature of the data at the individual subject level. We encourage researchers to consider the linear mixed model approach because it is more efficient, in that the whole data set is analysed in one step, is associated with less bias and results in greater statistical power compared with conventional analyses of baroreflex sensitivity for samples of subjects.",
author = "Greg Atkinson and Batterham, {Alan M.} and Helen Jones and Taylor, {Chloe E.} and Willie, {Chris K.} and Tzeng, {Yu Chieh}",
year = "2011",
month = "1",
day = "1",
doi = "10.1111/j.1475-097X.2010.00974.x",
language = "English",
volume = "31",
pages = "80--82",
journal = "Clinical Physiology and Functional Imaging",
issn = "1475-0961",
publisher = "Wiley-Blackwell Publishing Ltd",
number = "1",

}

Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity. / Atkinson, Greg; Batterham, Alan M.; Jones, Helen; Taylor, Chloe E.; Willie, Chris K.; Tzeng, Yu Chieh.

In: Clinical Physiology and Functional Imaging, Vol. 31, No. 1, 01.01.2011, p. 80-82.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Appropriate within-subjects statistical models for the analysis of baroreflex sensitivity

AU - Atkinson, Greg

AU - Batterham, Alan M.

AU - Jones, Helen

AU - Taylor, Chloe E.

AU - Willie, Chris K.

AU - Tzeng, Yu Chieh

PY - 2011/1/1

Y1 - 2011/1/1

N2 - An individual's baroreflex sensitivity is typically described by the relationship between serial manipulations in systolic blood pressure and the changes in pulse interval. Although this experimental approach is essentially within-subjects in nature, least squares regression (LSR) analysis is typically employed by researchers to derive sensitivity slopes (gains) for individual subjects. These individual gains are then pooled as summary measures for various samples or experimental conditions. We highlight that the underlying assumption for LSR of case independence is violated with such an approach, resulting in possible estimation biases and compromised statistical power. Using a typical data set, we introduce more appropriate analyses based on the linear mixed model, which takes into account the correlated nature of the data at the individual subject level. We encourage researchers to consider the linear mixed model approach because it is more efficient, in that the whole data set is analysed in one step, is associated with less bias and results in greater statistical power compared with conventional analyses of baroreflex sensitivity for samples of subjects.

AB - An individual's baroreflex sensitivity is typically described by the relationship between serial manipulations in systolic blood pressure and the changes in pulse interval. Although this experimental approach is essentially within-subjects in nature, least squares regression (LSR) analysis is typically employed by researchers to derive sensitivity slopes (gains) for individual subjects. These individual gains are then pooled as summary measures for various samples or experimental conditions. We highlight that the underlying assumption for LSR of case independence is violated with such an approach, resulting in possible estimation biases and compromised statistical power. Using a typical data set, we introduce more appropriate analyses based on the linear mixed model, which takes into account the correlated nature of the data at the individual subject level. We encourage researchers to consider the linear mixed model approach because it is more efficient, in that the whole data set is analysed in one step, is associated with less bias and results in greater statistical power compared with conventional analyses of baroreflex sensitivity for samples of subjects.

UR - http://www.scopus.com/inward/record.url?scp=78650114510&partnerID=8YFLogxK

U2 - 10.1111/j.1475-097X.2010.00974.x

DO - 10.1111/j.1475-097X.2010.00974.x

M3 - Article

VL - 31

SP - 80

EP - 82

JO - Clinical Physiology and Functional Imaging

JF - Clinical Physiology and Functional Imaging

SN - 1475-0961

IS - 1

ER -