Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Sample size and dimensionality in multivariate classification: implications for body surface potential mapping.

G Kozmann1, R L Lux, M Scott

  • 1Nora Eccles Harrison Cardiovascular Research and Training Institute, University of Utah, Salt Lake City.

Computers and Biomedical Research, an International Journal
|April 1, 1991
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Decision Support, Knowledge Representation and Management.

Yearbook of medical informatics·2016
Same author

Education in medical informatics on the basis of the information technology curriculum at the Veszprém University.

Yearbook of medical informatics·2016
Same author

Human brain distinctiveness based on EEG spectral coherence connectivity.

IEEE transactions on bio-medical engineering·2014
Same author

Statistics: more than pictures.

The Journal of small animal practice·2012
Same author

Gamow-Teller transition strengths from 56Ni.

Physical review letters·2011
Same author

Statistics: making sense of what we see.

The Journal of small animal practice·2011
Same journal

Wavelet domain nonlinear filtering for evoked potential signal enhancement.

Computers and biomedical research, an international journal·2001
Same journal

Time-frequency analysis of the RT and RR variability to stratify hypertrophic cardiomyopathy patients.

Computers and biomedical research, an international journal·2001
Same journal

Controlled auxotonic twitch in papillary muscle: a new computer-based control approach.

Computers and biomedical research, an international journal·2001
Same journal

Controlling for chance agreement in the validation of medical expert systems with no gold standard: PNEUMON-IA and RENOIR revisited.

Computers and biomedical research, an international journal·2001
Same journal

Artificial neural networks compared to factor analysis for low-dimensional classification of high-dimensional body fat topography data of healthy and diabetic subjects.

Computers and biomedical research, an international journal·2000
Same journal

On the parallelization of linkmap from the LINKAGE/FASTLINK package.

Computers and biomedical research, an international journal·2000
See all related articles

Determining the correct number of features for multivariate classification is crucial. This study provides guidelines to ensure reliable results, even with limited sample sizes in clinical body surface potential mapping studies.

Area of Science:

  • Statistics
  • Machine Learning
  • Biomedical Engineering

Background:

  • Multivariate classification studies require careful selection of features and adequate sample sizes for reliable results.
  • Existing guidelines for determining appropriate sample sizes and feature numbers are often insufficient, particularly in complex fields like clinical body surface potential mapping (BSPM).
  • Ensuring statistical reliability (Dmax < 0.2) is critical for valid group comparisons in these studies.

Purpose of the Study:

  • To establish empirically determined guidelines for selecting the optimal number of features in multivariate classification based on sample size.
  • To quantify sample size requirements for achieving a prescribed level of homogeneity between sample sets.
  • To assess the adequacy of sample sizes and feature selection in published clinical body surface potential mapping (BSPM) research.

Related Experiment Videos

Main Methods:

  • Investigated sample size adequacy by projecting sample sets onto the best separating direction and measuring the distance between them.
  • Quantified sample homogeneity as a function of dimensionality (M) and sample size (N) using maximum absolute distances (Dmax) between empirical cumulative probability distributions.
  • Utilized computer-generated datasets to estimate probability distributions and established an empirical relationship with the Kolmogorov limiting distribution L(z).

Main Results:

  • Established an empirical relationship between sample size, dimensionality, and the reliability of multivariate classification.
  • Analysis of 34 clinical body surface potential mapping (BSPM) papers revealed that 30% could only use one feature and 6% could use two features for reliable statistical group representation.
  • A significant 56% of published BSPM studies lacked sufficient sample sizes to guarantee reliability even for a single feature.

Conclusions:

  • Empirically derived guidelines are presented for specifying the number of features appropriate for multivariate classification studies given specific sample sizes.
  • The findings highlight a common inadequacy in sample sizes within published clinical body surface potential mapping (BSPM) research, compromising statistical reliability.
  • Researchers should carefully consider these guidelines to ensure the validity and reproducibility of their multivariate classification findings, particularly in biomedical applications.