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

Retro-regression--another important multivariate regression improvement.

M Randić1

  • 1Department of Mathematics and Computer Science, Drake University, Des Moines, Iowa 50311, USA.

Journal of Chemical Information and Computer Sciences
|June 21, 2001
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

Comparison of sequences as a method for evaluation of the molecular similarity.

Journal of computational chemistry·2017
Same author

On the history of the connectivity index: from the connectivity index to the exact solution of the protein alignment problem.

SAR and QSAR in environmental research·2015
Same author

Anthraquinone profile, antioxidant and antimicrobial activity of bark extracts of Rhamnus alaternus, R. fallax, R. intermedia and R. pumila.

Food chemistry·2012
Same author

On graphical representation of trans-membrane proteins.

SAR and QSAR in environmental research·2012
Same author

Very efficient search for protein alignment--VESPA.

Journal of computational chemistry·2012
Same author

Spectral representation of reduced protein models.

SAR and QSAR in environmental research·2009

Multivariate regression analysis (MRA) faces coefficient instability issues, termed "nightmares." A new retro-regression (RR) method resolves these unpredictable changes in descriptor selection for improved model reliability.

Area of Science:

  • Quantitative Chemistry
  • Statistical Modeling
  • Cheminformatics

Background:

  • Multivariate regression analysis (MRA) is susceptible to coefficient instability, known as the "nightmare of the first kind," where descriptor inclusion/exclusion causes unpredictable coefficient changes.
  • A more severe issue, the "nightmare of the second kind," occurs during optimal descriptor selection from large pools, leading to descriptor replacement during stepwise regression.

Purpose of the Study:

  • To address and resolve the coefficient instability problems in multivariate regression analysis (MRA).
  • To introduce a novel method, retro-regression (RR), capable of overcoming the "nightmares" of MRA.

Main Methods:

  • Review of coefficient instabilities in MRA, specifically the "nightmare of the first kind" (descriptor inclusion/exclusion) and "nightmare of the second kind" (optimal descriptor selection).

Related Experiment Videos

  • Illustration of a new procedure using boiling points of nonanes with ordered connectivity, greedy algorithm, and exhaustive search for descriptor ordering.
  • Outline of retro-regression (RR), a novel variant of MRA designed to resolve ambiguities.

    • Main Results:

    • The study identifies and defines two major problems in MRA: coefficient instability upon descriptor changes and descriptor replacement during selection.
    • A new procedure is presented and illustrated using nonane boiling points, demonstrating its application with different descriptor ordering strategies.
    • Retro-regression (RR) is shown to effectively resolve the ambiguities associated with both types of MRA nightmares.

    Conclusions:

    • Coefficient instabilities in MRA ('nightmares') pose significant challenges in regression modeling.
    • The proposed retro-regression (RR) method offers a robust solution to these MRA ambiguities.
    • RR enhances the reliability and predictability of regression models, particularly in descriptor selection processes.