Loading…

Instability of ordination results under changes in input data order: explanations and remedies

Correspondence analysis (CA) and its Detrended form (DCA) produced by the program CANOCO are unstable under reordering of the species and sites in the input data matrix. In CA, the main cause of the instability is the use of insufficiently stringent convergence criteria in the power algorithm used t...

Full description

Saved in:
Bibliographic Details
Published in:Journal of vegetation science 1997-06, Vol.8 (3), p.447-454
Main Authors: Oksanen, Jari, Minchin, Peter R.
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853
cites cdi_FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853
container_end_page 454
container_issue 3
container_start_page 447
container_title Journal of vegetation science
container_volume 8
creator Oksanen, Jari
Minchin, Peter R.
description Correspondence analysis (CA) and its Detrended form (DCA) produced by the program CANOCO are unstable under reordering of the species and sites in the input data matrix. In CA, the main cause of the instability is the use of insufficiently stringent convergence criteria in the power algorithm used to estimate the eigenvalues. The use of stricter criteria gives results that are acceptably stable. The divisive classification program TWINSPAN uses CA based on a similar algorithm, but with extremely lax convergence criteria, and is thus susceptible to extreme instability. We detected an order-dependent programming error in the non-linear rescaling procedure that forms part of DCA. When this bug is corrected, much of the instability in DCA disappears. The stability of DCA solutions is further enhanced by the use of strict convergence criteria. In our trials, much of the instability occurred in axes 3 and 4, but ine should not assume that published two-dimensional ordinations are sufficiently accurate. Data sets which have pairs of almost equal eigenvalues among the first three axes could suffer from marked instability in the first two dimensions. We recommend that a debugged, strict version of CANOCO be released. Meanwhile, users can check the stability of their CA and DCA ordinations using the software that we have made available on the World Wide Web (http:// www.helsinki.fi/~jhoksane/). An accurate program for CA, a debugged, strict version of DECORANA (for DCA) and a strict version of TWINSPAN are also available at our site.
doi_str_mv 10.2307/3237336
format article
fullrecord <record><control><sourceid>jstor_cross</sourceid><recordid>TN_cdi_crossref_primary_10_2307_3237336</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><jstor_id>3237336</jstor_id><sourcerecordid>3237336</sourcerecordid><originalsourceid>FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853</originalsourceid><addsrcrecordid>eNp1kFFLwzAUhYMoOKf4BwTz5oNUk9w26XyToXMy9WFugg-GtElmZteOpMPt39tRmU_ChXvgfOfCPQidUnLFgIhrYCAA-B7qUJ7EEaUE9htNCYl6DOAQHYUwJ4SKHqcd9DEsQ60yV7h6gyuLK69dqWpXldibsCrqgFelNh7nn6qcmYBd2cxyVWOtarXFjb_BZr0sVBsLWJW6yS6MdiYcowOrimBOfncXTe7vXvsP0ehlMOzfjqIcYiYixVROBHDCewAiTbTNRJwqQ0VOc2XBEpHZmDWCa5IlmU20BpZqbeI0o2kCXXTR3s19FYI3Vi69Wyi_kZTIbS3yt5aGvGzJb1eYzX-YfJyO00Z10VlLz0Nd-R39dyxqbRdqs97Zyn9JLkAk8u15IKcc-Pv4iUho-POWt6qSauZdkJMxIxQIS5uHezH8AJUdhIo</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Instability of ordination results under changes in input data order: explanations and remedies</title><source>Wiley:Jisc Collections:Wiley Read and Publish Open Access 2024-2025 (reading list)</source><source>JSTOR Archival Journals and Primary Sources Collection</source><creator>Oksanen, Jari ; Minchin, Peter R.</creator><creatorcontrib>Oksanen, Jari ; Minchin, Peter R.</creatorcontrib><description>Correspondence analysis (CA) and its Detrended form (DCA) produced by the program CANOCO are unstable under reordering of the species and sites in the input data matrix. In CA, the main cause of the instability is the use of insufficiently stringent convergence criteria in the power algorithm used to estimate the eigenvalues. The use of stricter criteria gives results that are acceptably stable. The divisive classification program TWINSPAN uses CA based on a similar algorithm, but with extremely lax convergence criteria, and is thus susceptible to extreme instability. We detected an order-dependent programming error in the non-linear rescaling procedure that forms part of DCA. When this bug is corrected, much of the instability in DCA disappears. The stability of DCA solutions is further enhanced by the use of strict convergence criteria. In our trials, much of the instability occurred in axes 3 and 4, but ine should not assume that published two-dimensional ordinations are sufficiently accurate. Data sets which have pairs of almost equal eigenvalues among the first three axes could suffer from marked instability in the first two dimensions. We recommend that a debugged, strict version of CANOCO be released. Meanwhile, users can check the stability of their CA and DCA ordinations using the software that we have made available on the World Wide Web (http:// www.helsinki.fi/~jhoksane/). An accurate program for CA, a debugged, strict version of DECORANA (for DCA) and a strict version of TWINSPAN are also available at our site.</description><identifier>ISSN: 1100-9233</identifier><identifier>EISSN: 1654-1103</identifier><identifier>DOI: 10.2307/3237336</identifier><language>eng</language><publisher>Oxford, UK: Blackwell Publishing Ltd</publisher><subject>Algorithm ; Algorithms ; cluster analysis ; Clustering ; Computer software ; correlation analysis ; Correspondence analysis ; Datasets ; Detrended correspondence analysis ; Eigenanalysis ; Eigenvalues ; Eigenvectors ; equations ; Fortran ; Forum ; Input data ; mathematical models ; Non-linear rescaling ; Ordination ; plant ecology ; Polls ; Software bugs ; statistical analysis ; theory ; Tolerance</subject><ispartof>Journal of vegetation science, 1997-06, Vol.8 (3), p.447-454</ispartof><rights>Copyright 1997 IAVS; Opulus Press Uppsala</rights><rights>1997 IAVS ‐ the International Association of Vegetation Science</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853</citedby><cites>FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://www.jstor.org/stable/pdf/3237336$$EPDF$$P50$$Gjstor$$H</linktopdf><linktohtml>$$Uhttps://www.jstor.org/stable/3237336$$EHTML$$P50$$Gjstor$$H</linktohtml><link.rule.ids>314,780,784,27923,27924,58237,58470</link.rule.ids></links><search><creatorcontrib>Oksanen, Jari</creatorcontrib><creatorcontrib>Minchin, Peter R.</creatorcontrib><title>Instability of ordination results under changes in input data order: explanations and remedies</title><title>Journal of vegetation science</title><description>Correspondence analysis (CA) and its Detrended form (DCA) produced by the program CANOCO are unstable under reordering of the species and sites in the input data matrix. In CA, the main cause of the instability is the use of insufficiently stringent convergence criteria in the power algorithm used to estimate the eigenvalues. The use of stricter criteria gives results that are acceptably stable. The divisive classification program TWINSPAN uses CA based on a similar algorithm, but with extremely lax convergence criteria, and is thus susceptible to extreme instability. We detected an order-dependent programming error in the non-linear rescaling procedure that forms part of DCA. When this bug is corrected, much of the instability in DCA disappears. The stability of DCA solutions is further enhanced by the use of strict convergence criteria. In our trials, much of the instability occurred in axes 3 and 4, but ine should not assume that published two-dimensional ordinations are sufficiently accurate. Data sets which have pairs of almost equal eigenvalues among the first three axes could suffer from marked instability in the first two dimensions. We recommend that a debugged, strict version of CANOCO be released. Meanwhile, users can check the stability of their CA and DCA ordinations using the software that we have made available on the World Wide Web (http:// www.helsinki.fi/~jhoksane/). An accurate program for CA, a debugged, strict version of DECORANA (for DCA) and a strict version of TWINSPAN are also available at our site.</description><subject>Algorithm</subject><subject>Algorithms</subject><subject>cluster analysis</subject><subject>Clustering</subject><subject>Computer software</subject><subject>correlation analysis</subject><subject>Correspondence analysis</subject><subject>Datasets</subject><subject>Detrended correspondence analysis</subject><subject>Eigenanalysis</subject><subject>Eigenvalues</subject><subject>Eigenvectors</subject><subject>equations</subject><subject>Fortran</subject><subject>Forum</subject><subject>Input data</subject><subject>mathematical models</subject><subject>Non-linear rescaling</subject><subject>Ordination</subject><subject>plant ecology</subject><subject>Polls</subject><subject>Software bugs</subject><subject>statistical analysis</subject><subject>theory</subject><subject>Tolerance</subject><issn>1100-9233</issn><issn>1654-1103</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>1997</creationdate><recordtype>article</recordtype><recordid>eNp1kFFLwzAUhYMoOKf4BwTz5oNUk9w26XyToXMy9WFugg-GtElmZteOpMPt39tRmU_ChXvgfOfCPQidUnLFgIhrYCAA-B7qUJ7EEaUE9htNCYl6DOAQHYUwJ4SKHqcd9DEsQ60yV7h6gyuLK69dqWpXldibsCrqgFelNh7nn6qcmYBd2cxyVWOtarXFjb_BZr0sVBsLWJW6yS6MdiYcowOrimBOfncXTe7vXvsP0ehlMOzfjqIcYiYixVROBHDCewAiTbTNRJwqQ0VOc2XBEpHZmDWCa5IlmU20BpZqbeI0o2kCXXTR3s19FYI3Vi69Wyi_kZTIbS3yt5aGvGzJb1eYzX-YfJyO00Z10VlLz0Nd-R39dyxqbRdqs97Zyn9JLkAk8u15IKcc-Pv4iUho-POWt6qSauZdkJMxIxQIS5uHezH8AJUdhIo</recordid><startdate>199706</startdate><enddate>199706</enddate><creator>Oksanen, Jari</creator><creator>Minchin, Peter R.</creator><general>Blackwell Publishing Ltd</general><general>Opulus Press</general><scope>FBQ</scope><scope>BSCLL</scope><scope>AAYXX</scope><scope>CITATION</scope></search><sort><creationdate>199706</creationdate><title>Instability of ordination results under changes in input data order: explanations and remedies</title><author>Oksanen, Jari ; Minchin, Peter R.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>1997</creationdate><topic>Algorithm</topic><topic>Algorithms</topic><topic>cluster analysis</topic><topic>Clustering</topic><topic>Computer software</topic><topic>correlation analysis</topic><topic>Correspondence analysis</topic><topic>Datasets</topic><topic>Detrended correspondence analysis</topic><topic>Eigenanalysis</topic><topic>Eigenvalues</topic><topic>Eigenvectors</topic><topic>equations</topic><topic>Fortran</topic><topic>Forum</topic><topic>Input data</topic><topic>mathematical models</topic><topic>Non-linear rescaling</topic><topic>Ordination</topic><topic>plant ecology</topic><topic>Polls</topic><topic>Software bugs</topic><topic>statistical analysis</topic><topic>theory</topic><topic>Tolerance</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Oksanen, Jari</creatorcontrib><creatorcontrib>Minchin, Peter R.</creatorcontrib><collection>AGRIS</collection><collection>Istex</collection><collection>CrossRef</collection><jtitle>Journal of vegetation science</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Oksanen, Jari</au><au>Minchin, Peter R.</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Instability of ordination results under changes in input data order: explanations and remedies</atitle><jtitle>Journal of vegetation science</jtitle><date>1997-06</date><risdate>1997</risdate><volume>8</volume><issue>3</issue><spage>447</spage><epage>454</epage><pages>447-454</pages><issn>1100-9233</issn><eissn>1654-1103</eissn><abstract>Correspondence analysis (CA) and its Detrended form (DCA) produced by the program CANOCO are unstable under reordering of the species and sites in the input data matrix. In CA, the main cause of the instability is the use of insufficiently stringent convergence criteria in the power algorithm used to estimate the eigenvalues. The use of stricter criteria gives results that are acceptably stable. The divisive classification program TWINSPAN uses CA based on a similar algorithm, but with extremely lax convergence criteria, and is thus susceptible to extreme instability. We detected an order-dependent programming error in the non-linear rescaling procedure that forms part of DCA. When this bug is corrected, much of the instability in DCA disappears. The stability of DCA solutions is further enhanced by the use of strict convergence criteria. In our trials, much of the instability occurred in axes 3 and 4, but ine should not assume that published two-dimensional ordinations are sufficiently accurate. Data sets which have pairs of almost equal eigenvalues among the first three axes could suffer from marked instability in the first two dimensions. We recommend that a debugged, strict version of CANOCO be released. Meanwhile, users can check the stability of their CA and DCA ordinations using the software that we have made available on the World Wide Web (http:// www.helsinki.fi/~jhoksane/). An accurate program for CA, a debugged, strict version of DECORANA (for DCA) and a strict version of TWINSPAN are also available at our site.</abstract><cop>Oxford, UK</cop><pub>Blackwell Publishing Ltd</pub><doi>10.2307/3237336</doi><tpages>8</tpages></addata></record>
fulltext fulltext
identifier ISSN: 1100-9233
ispartof Journal of vegetation science, 1997-06, Vol.8 (3), p.447-454
issn 1100-9233
1654-1103
language eng
recordid cdi_crossref_primary_10_2307_3237336
source Wiley:Jisc Collections:Wiley Read and Publish Open Access 2024-2025 (reading list); JSTOR Archival Journals and Primary Sources Collection
subjects Algorithm
Algorithms
cluster analysis
Clustering
Computer software
correlation analysis
Correspondence analysis
Datasets
Detrended correspondence analysis
Eigenanalysis
Eigenvalues
Eigenvectors
equations
Fortran
Forum
Input data
mathematical models
Non-linear rescaling
Ordination
plant ecology
Polls
Software bugs
statistical analysis
theory
Tolerance
title Instability of ordination results under changes in input data order: explanations and remedies
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-10T10%3A40%3A27IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-jstor_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Instability%20of%20ordination%20results%20under%20changes%20in%20input%20data%20order:%20explanations%20and%20remedies&rft.jtitle=Journal%20of%20vegetation%20science&rft.au=Oksanen,%20Jari&rft.date=1997-06&rft.volume=8&rft.issue=3&rft.spage=447&rft.epage=454&rft.pages=447-454&rft.issn=1100-9233&rft.eissn=1654-1103&rft_id=info:doi/10.2307/3237336&rft_dat=%3Cjstor_cross%3E3237336%3C/jstor_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c3427-a2ac073606933785dfb748ae17c1caf3f07bf42f3f6d0b5bf5dd328dde48b1853%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_jstor_id=3237336&rfr_iscdi=true