Five myths about variable selection.

Georg Heinze, Daniela Dunkler

January 2017 Transpl Int

Synopsis of Social media discussions

The group collectively recognizes the article's critique of common myths, such as univariate screening, and discusses alternatives supported by multiple references, reflecting a thorough understanding and a desire to improve research methodology. Words like 'bad idea', 'not good', and references to authoritative papers highlight the critical and impactful tone of the discussions.

A
Agreement
 Strong agreement

Most discussions strongly concur with the article, emphasizing that traditional variable selection methods like univariate screening are often problematic and should be avoided.

I
Interest
 High level of interest

Participants demonstrate high interest by referencing multiple detailed papers, formulas, and sharing insights about the myths surrounding variable selection.

E
Engagement
 High engagement

There is deep engagement evidenced by referencing specific studies, debating methodological nuances, and sharing personal opinions on best practices.

I
Impact
 High level of impact

The discussions have high impact as they seem to influence research practices, with many advocating for rethinking standard variable selection strategies in applied research.

Social Mentions

YouTube

2 Videos

Twitter

141 Posts

News

2 Articles

Metrics

Video Views

231

Total Likes

580

Extended Reach

708,313

Social Features

145

Timeline: Posts about article

Top Social Media Posts

Posts referencing the article

Improving Variable Selection Methods in Epidemiological Research

Improving Variable Selection Methods in Epidemiological Research

This lecture provides an overview of recent reviews on variable selection in epidemiology, highlighting ongoing reliance on traditional methods like stepwise analysis and advocating for improved practices using causal diagrams and prior knowledge.

Jochen Kruppa-Scheetz

May 10, 2021

144 views

Improving Variable Selection Methods in Epidemiological Research

Improving Variable Selection Methods in Epidemiological Research

This lecture discusses the importance of referencing when selecting variables, highlighting current practices and potential improvements in epidemiological variable selection methods.

Jochen Kruppa-Scheetz

May 10, 2021

87 views

  • MITTI
    @MITTI12101 (Twitter)

    RT @hommedefer3: @naka_takaya Steyerberg先生の本でも明確に避けるべきとは書かれていないように思いますが、ステップワイズと同じ問題があることは明記されてますね。 このレビュー↓でも予測観点でも単変量スクリーニングは推奨されてないです。 h…
    view full post

    February 14, 2025

    1

  • ironman
    @hommedefer3 (Twitter)

    @naka_takaya Steyerberg先生の本でも明確に避けるべきとは書かれていないように思いますが、ステップワイズと同じ問題があることは明記されてますね。 このレビュー↓でも予測観点でも単変量スクリーニングは推奨されてないです。 https://t.co/x73dLhKHvC そこで引用されてる文献↓ https://t.co/HemVvDxsvz
    view full post

    February 13, 2025

    24

    1

  • Reaper001
    @ShadesOfReaper (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    October 12, 2024

    9

  • nissinbo
    @nissinbo_ (Twitter)

    @CareerPharm ここらへんですかね？ https://t.co/pfnpS1qgLl https://t.co/9kbSpUrwiq
    view full post

    June 18, 2024

  • Mari Takashima
    @MariDTakashima (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    May 16, 2024

    40

  • Huaiyu Zang
    @HuaiyuZang (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 6, 2024

    40

  • Pedro Lopez-Ayala
    @LopezAyalaP (Twitter)

    @manuel_carnero @cirujano_cardio Este paper es una buena lectura https://t.co/JPutqDKZCr El libro de referencia para prediction modeling es "regression modeling strategies" de Harrell, y este capítulo es un buen comienzo. https://t.co/i8qtcHnQk5
    view full post

    September 6, 2023

    1

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @andersskyrud @trumanfrancis @statsepi It's well know this is problematic https://t.co/qJiQHkvKOH
    view full post

    November 14, 2022

    4

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @HelenFry1 @Donato29708932 @Richard_D_Riley Please find here two further refs: https://t.co/ZGDjtQ9Hfm And https://t.co/NloWPm7Lr9
    view full post

    October 29, 2022

  • Leonardo Furlan
    @LeoFurlanPhD (Twitter)

    @Georg__Heinze, just read your Review. Which is preferable? 1) full model, remove all insignificant predictors (sequentially or all at once? which cut-off?) until only significant predictors are left 2) full model and select predictors using P < 0.157 https://t.co/feIAwn6J4A
    view full post

    September 1, 2022

  • José A. Calvache
    @jacalvache (Twitter)

    RT @Georg__Heinze: @fathus44 Basically explained here: https://t.co/NloWPlQazB
    view full post

    April 17, 2022

    1

  • NephrologyLeuven
    @NephroLeuven (Twitter)

    RT @MaartenvSmeden: Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve an…
    view full post

    April 1, 2022

    5

  • Yu Sun Bin
    @yusunbin (Twitter)

    RT @MaartenvSmeden: @JanFWeller @JustinWilkins Yes: https://t.co/qJiQHkvKOH
    view full post

    March 27, 2022

    5

  • MK
    @muralijnu1 (Twitter)

    RT @MaartenvSmeden: @JanFWeller @JustinWilkins Yes: https://t.co/qJiQHkvKOH
    view full post

    March 26, 2022

    5

  • Carlos E Lourenco
    @caerib (Twitter)

    RT @MaartenvSmeden: @JanFWeller @JustinWilkins Yes: https://t.co/qJiQHkvKOH
    view full post

    March 26, 2022

    5

  • @docbio.bsky.social
    @DocBio1509 (Twitter)

    RT @MaartenvSmeden: @JanFWeller @JustinWilkins Yes: https://t.co/qJiQHkvKOH
    view full post

    March 26, 2022

    5

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @JanFWeller @JustinWilkins Yes: https://t.co/qJiQHkvKOH
    view full post

    March 26, 2022

    11

    5

  • kokako
    @kokako65541706 (Twitter)

    RT @MaartenvSmeden: Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve an…
    view full post

    March 18, 2022

    5

  • Gregor Bond
    @BondGregor (Twitter)

    RT @MaartenvSmeden: Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve an…
    view full post

    March 17, 2022

    5

  • Zedd
    @TrdisOprtr (Twitter)

    RT @MaartenvSmeden: Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve an…
    view full post

    March 17, 2022

    5

  • Maarten Naesens
    @mnaesens (Twitter)

    RT @MaartenvSmeden: Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve an…
    view full post

    March 17, 2022

    5

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    Data driven variable selection - unlike popular views, removing variables that are "not significant" may not improve analyses https://t.co/qJiQHkvKOH https://t.co/OmENQ1msTc
    view full post

    March 17, 2022

    27

    5

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @fathus44 Basically explained here: https://t.co/NloWPlQazB
    view full post

    March 9, 2022

    1

    1

  • Laure Wynants @laurewynants.bsky.social
    @laure_wynants (Twitter)

    @EpiSconroy https://t.co/aRg6MNvFCU
    view full post

    February 2, 2022

    3

  • Scott Parker
    @ParkerS15013836 (Twitter)

    RT @MaartenvSmeden: @alwynkotzee @mababyak @trumanfrancis https://t.co/qJiQHkvKOH
    view full post

    November 17, 2021

    1

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @alwynkotzee @mababyak @trumanfrancis https://t.co/qJiQHkvKOH
    view full post

    November 17, 2021

    5

    1

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @DrBenMoran @load_dependent @iceman_ex Hear hear. And even *if* you think variable selection is a good idea, please forget about univariable selection https://t.co/qJiQHkNlGf
    view full post

    September 21, 2021

    4

  • Bastien Rioux
    @B_Rioux (Twitter)

    RT @stroke_methods: Variable selection approaches are often employed in stroke research. Great article below on five myths.⭐️ Myth #1: Onl…
    view full post

    May 31, 2021

    1

  • Stroke Epidemiology Methods
    @stroke_methods (Twitter)

    Variable selection approaches are often employed in stroke research. Great article below on five myths.⭐️ Myth #1: Only variables with proven univariable-model significance should be included in a model ❌ Read more here⬇️ https://t.co/odMkGEElR6
    view full post

    May 31, 2021

    1

  • Anne-Laure Boulesteix
    @BoulesteixLaure (Twitter)

    RT @Georg__Heinze: @MaartenvSmeden https://t.co/ChwL3iGk62
    view full post

    May 20, 2021

    1

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @MaartenvSmeden https://t.co/ChwL3iGk62
    view full post

    May 20, 2021

    16

    1

  • tsubameuni
    @fjge3dkgutcka (Twitter)

    RT @triadsou: @Shuntarooo3 Five myths about variable selection. Georg Heinze, Daniela Dunkler. https://t.co/BS6VkqIci9
    view full post

    April 16, 2021

    1

  • Triad sou.
    @triadsou (Twitter)

    @Shuntarooo3 Five myths about variable selection. Georg Heinze, Daniela Dunkler. https://t.co/BS6VkqIci9
    view full post

    April 16, 2021

    21

    1

  • Andrew Yockey, PhD (He/Him)
    @YockeyAndrew (Twitter)

    Five myths about variable selection https://t.co/GWOvLNzQkp
    view full post

    April 12, 2021

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @JLRohmann Depends on field. Medical/epi papers: https://t.co/NloWPlQazB Statistical papers: https://t.co/NloWPlQazB
    view full post

    March 26, 2021

    2

  • Zoubir Djerada
    @ZD79055334 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 21, 2021

    40

  • Charlse Newman: Official War Artist, Possum Party
    @narngaloo (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 21, 2021

    9

  • Yukiko Ezure
    @Yukie40899965 (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 20, 2021

    9

  • Luis Fernando Gómez Gutiérrez.
    @LuisFernandoGm6 (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 20, 2021

    9

  • Ramiro Bermudez-I.MD
    @davidbermudezi (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 20, 2021

    9

  • Daniel P. Moriarity
    @dp_moriarity (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 20, 2021

    9

  • Carlos Rubén
    @caruben89 (Twitter)

    RT @MaartenvSmeden: 1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "…
    view full post

    March 20, 2021

    9

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    1) As a precursor to multivariable analyses, the associations between each individual covariate and outcome are often "screened" for significance. Often does more harm than good, so don't bother doing or worry about it https://t.co/fR6fN2HkeM
    view full post

    March 20, 2021

    73

    9

  • Kareem El-Fass BSc, PharmD, MSc
    @kareemelfass (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 15, 2021

    40

  • Danilo Moggia, Ph.D.
    @psiconstructiva (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 11, 2021

    40

  • owela ngezihlangu zamadoda ✨
    @Sanda_Maz (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 11, 2021

    40

  • Jinn-Yuh Guh
    @guhjy (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 11, 2021

    40

  • Thomas Speidel
    @ThomasSpeidel (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Dr. Maryam A Nejad
    @MaryamANEJAD (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Sun-Ho Lee MD PhD
    @SunHoLee15 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • CaTaLiNA Research Initiative
    @CaTaLiNA_Ecu (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Edouard L. Fu
    @FuEdouard (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Garrett Bullock PT, DPT, DPhil
    @DrGSBullock (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Andrew Yockey, PhD (He/Him)
    @YockeyAndrew (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Brainheartnexus
    @brainheartnexus (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • rafael mello galliez
    @RafaelGalliez (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Jessica Laine Carmeli, PhD
    @Laine_J_E_ (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Diego Costa
    @drdiegocosta (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Atiq Ur Rehman
    @atiqmarwat12 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Lars L. Andersen
    @LarsLAndersen (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Kazuki Yoshida
    @kaz_yos (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Mari Takashima
    @MariDTakashima (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40


  • @johnthomas75 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Martin B Whyte
    @mbwhyte1 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Hanis A K
    @90sKeed (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Ignacio Borlaf-Mena
    @borlafgis (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Matt Craddock @mattcraddock.bsky.social
    @Matt_Craddock (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Gerd Antes
    @gerdantes (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Miquel Bennasar Veny
    @miquelbennasar (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Oscar Ponce
    @PonceOJ (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Chomes
    @CHOMES102 (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Serius Black
    @dnameis_paone (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    March 1, 2021

    40

  • Roger Hilfiker
    @RogerHilfiker (Twitter)

    RT @MaartenvSmeden: @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2)…
    view full post

    February 28, 2021

    40

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @coenen_pieter 1) univariate screening is a very, very bad idea in almost all circumstances: https://t.co/qJiQHkvKOH 2) stepwise too, for better guidance see: https://t.co/AZ2CjbllVH 3) with multivariate people typically mean to say multivariable
    view full post

    February 28, 2021

    117

    40

  • Assumpta Nantume
    @Ansolome (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 18, 2020

    13

  • Ettore Ambrosini
    @EttAmbros (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 18, 2020

    13

  • Mir Henglin
    @_mir_cat (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Alessio Gerussi
    @AlessioGerussi (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Marije van der Lee
    @Marijeliesbeth1 (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Athanasios Kousathanas
    @akousathanas (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Serius Black
    @dnameis_paone (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Jesper Kivelä
    @JesperKivela (Twitter)

    @MaartenvSmeden 5 myths with NO! according to @Georg__Heinze "Often there is no scientific reason to perform variable selection." https://t.co/Ifvzcfx3rs
    view full post

    November 17, 2020

    2

  • Letícia Raposo
    @Leticia_Raposo (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • James Pitt
    @Sahelanth (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Roger Hilfiker
    @RogerHilfiker (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Dale Needham, MD, PhD
    @DrDaleNeedham (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Ian Hussey
    @ianhussey (Twitter)

    RT @MaartenvSmeden: Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoid…
    view full post

    November 17, 2020

    13

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    Given the interim results and questions, I think it is worthwhile to point out that univariate screening is best *avoided* for nearly all applications imaginable, e.g. see: https://t.co/qJiQHkvKOH https://t.co/AZ2CjbllVH https://t.co/mznJoAWzi2
    view full post

    November 17, 2020

    59

    13

  • Saurab Sharma, PhD
    @link_physio (Twitter)

    RT @GSCollins: @danrhon @TenanATC @Nrupen @DrGSBullock Two good papers on the topic of variable selection that discuss many issues (includi…
    view full post

    November 10, 2020

    3

  • Didier Brassard, PhD
    @didier_dtp (Twitter)

    RT @MaartenvSmeden: @GSCollins @danrhon @TenanATC @Nrupen @DrGSBullock I would add this nice and more recent article by @Georg__Heinze un…
    view full post

    November 9, 2020

    4

  • Garrett Bullock PT, DPT, DPhil
    @DrGSBullock (Twitter)

    RT @MaartenvSmeden: @GSCollins @danrhon @TenanATC @Nrupen @DrGSBullock I would add this nice and more recent article by @Georg__Heinze un…
    view full post

    November 9, 2020

    4

  • Dan Rhon
    @danrhon (Twitter)

    RT @MaartenvSmeden: @GSCollins @danrhon @TenanATC @Nrupen @DrGSBullock I would add this nice and more recent article by @Georg__Heinze un…
    view full post

    November 9, 2020

    4

  • Dan Rhon
    @danrhon (Twitter)

    RT @GSCollins: @danrhon @TenanATC @Nrupen @DrGSBullock Two good papers on the topic of variable selection that discuss many issues (includi…
    view full post

    November 9, 2020

    3

  • Fredrik Olsen
    @Olsefred (Twitter)

    RT @GSCollins: @danrhon @TenanATC @Nrupen @DrGSBullock Two good papers on the topic of variable selection that discuss many issues (includi…
    view full post

    November 9, 2020

    3

  • Georg Heinze
    @Georg__Heinze (Twitter)

    RT @MaartenvSmeden: @GSCollins @danrhon @TenanATC @Nrupen @DrGSBullock I would add this nice and more recent article by @Georg__Heinze un…
    view full post

    November 9, 2020

    4

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @GSCollins @danrhon @TenanATC @Nrupen @DrGSBullock I would add this nice and more recent article by @Georg__Heinze univariable selection: not providing benefits at best, destroys the analysis at worst https://t.co/fR6fN2HkeM https://t.co/MvlMyCaiI5
    view full post

    November 9, 2020

    15

    4

  • Gary Collins
    @GSCollins (Twitter)

    @danrhon @TenanATC @Nrupen @DrGSBullock Two good papers on the topic of variable selection that discuss many issues (including variable screening) are https://t.co/9iF9BcrZ9c and https://t.co/Uq6s4lhBy1
    view full post

    November 9, 2020

    9

    3

  • Robert Ancuceanu
    @Rancuceanu (Twitter)

    RT @Richard_D_Riley: Other refs: Sun et al. (1996). Inappropriate use of bivariable analysis to screen risk factors for use in multivariabl…
    view full post

    October 20, 2020

    4

  • murapo
    @murapo1 (Twitter)

    RT @Richard_D_Riley: Other refs: Sun et al. (1996). Inappropriate use of bivariable analysis to screen risk factors for use in multivariabl…
    view full post

    October 19, 2020

    4

  • Ram Bajpai
    @RamBajpai16 (Twitter)

    RT @Richard_D_Riley: Other refs: Sun et al. (1996). Inappropriate use of bivariable analysis to screen risk factors for use in multivariabl…
    view full post

    October 19, 2020

    4

  • Giuseppe Biondi-Zoccai
    @gbiondizoccai (Twitter)

    RT @Richard_D_Riley: Other refs: Sun et al. (1996). Inappropriate use of bivariable analysis to screen risk factors for use in multivariabl…
    view full post

    October 19, 2020

    4

  • Richard Riley (R²)
    @Richard_D_Riley (Twitter)

    Other refs: Sun et al. (1996). Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis. Journal of Clinical Epidemiology, 49, 907–916 Heinze & Dunkler. (2017). Five myths about variable selection. Transplant International, 30, 6–10.
    view full post

    October 19, 2020

    9

    4

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @fracardo @MedUni_Wien And as a starter I propose this one: https://t.co/ChwL3iGk62
    view full post

    July 1, 2020

    1

  • Zach Hopkins, MD
    @ZachHopkinsMD (Twitter)

    RT @MaartenvSmeden: @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) conf…
    view full post

    June 6, 2020

    5

  • Serius Black
    @dnameis_paone (Twitter)

    RT @MaartenvSmeden: @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) conf…
    view full post

    June 2, 2020

    5

  • John Hodsoll
    @HodsollJohn (Twitter)

    RT @MaartenvSmeden: @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) conf…
    view full post

    June 2, 2020

    5

  • Abraham E. Gracia R.
    @Abraham_RMI (Twitter)

    RT @MaartenvSmeden: @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) conf…
    view full post

    June 2, 2020

    5

  • Chris Sheridan
    @sheridch90 (Twitter)

    RT @MaartenvSmeden: @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) conf…
    view full post

    June 2, 2020

    5

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @SoniaBoender @YourPaperSucks @elizabethsmout 1) univariable selection = not good idea: https://t.co/qJiQHkvKOH 2) confounder selection is complicated: https://t.co/bqgFN2udlz 3) interpreting anything after variable selection is complicated: https://t.co/5Sye6cIcD2
    view full post

    June 2, 2020

    17

    5

  • Ali Khaki
    @arkhaki (Twitter)

    @thenasheffect @ASCO @PGrivasMDPhD @uwurology @SeattleCCA @fredhutch @LDiamantMD @bilenma @sonpavde @MattGalsky @MonikaJoshimd @PedroIsaacsson Thank you! Re your q’s: 1. p<0.2 for univariate and 0.05 for mva. I don’t think there is a strict rule of p value for variable selection - see myth 5 here - https://t.co/h78KQ8gK5A 2. NLR and ANC were not included together. Table 3 in poster shows me testing both ANC and NLR.
    view full post

    February 14, 2020

    1

  • Matthew Vowels
    @matt_vowels (Twitter)

    the number of social science professors and grad students we see doing and encouraging this... #tryagainpsych https://t.co/Tj3zbbBAJf
    view full post

    September 10, 2019

    1

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @PhDemetri It means your reviewer believes in at least one or two myths about variable selection, unfortunately! https://t.co/R6Dwjt3CSn
    view full post

    September 10, 2019

    5

  • Stefan
    @stefjerofejew (Twitter)

    @PhDemetri See Myth #2 in this paper in easy medical practitioner language: https://t.co/SNpQppz0qL maybe helps
    view full post

    September 9, 2019

    4

  • Stephen John Senn
    @stephensenn (Twitter)

    RT @laure_wynants: @ChristosArgyrop @f2harrell @stephensenn @ADAlthousePhD @Georg__Heinze has written it down for applied researchers and m…
    view full post

    August 25, 2019

    2

  • Christian Staerk
    @StaerkChristian (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 8, 2019

    12

  • -
    @python_ds (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 3, 2019

    12

  • Hucello Chuyucello, PhD
    @MrFuntactic (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • bo ko
    @petmolle (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • John Boylan
    @jboylan (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Daniela Moga
    @epidmoga (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Chrissie
    @ChrisC_RN (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Erin Delker
    @erin_delker (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Domenico Bellomo
    @BellomoDomenico (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Pilar Garcia-Iglesias
    @Amleto3333 (Twitter)

    RT @MaartenvSmeden: @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf…
    view full post

    August 2, 2019

    12

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @ADAlthousePhD I really like these four papers: https://t.co/bqgFN2LOd7 https://t.co/HFtgA2RGAj https://t.co/AZ2CjbCWNf https://t.co/fR6fN2YV6k https://t.co/NSG7srb7xL
    view full post

    August 2, 2019

    82

    12

  • Sandro Rodriguez
    @Sandroppa (Twitter)

    RT @MaartenvSmeden: @FustoloSusanna This, for instance: https://t.co/fR6fN2HkeM https://t.co/xHb2vyjLvm
    view full post

    June 17, 2019

    3

  • Gabriel González Medina
    @Gabriel00602805 (Twitter)

    RT @MaartenvSmeden: @FustoloSusanna This, for instance: https://t.co/fR6fN2HkeM https://t.co/xHb2vyjLvm
    view full post

    June 17, 2019

    3

  • George Savva
    @georgemsavva (Twitter)

    RT @MaartenvSmeden: @FustoloSusanna This, for instance: https://t.co/fR6fN2HkeM https://t.co/xHb2vyjLvm
    view full post

    June 17, 2019

    3

  • Valirie Agbor
    @ValirieAgbor (Twitter)

    RT @JFronczekMD: @ValirieAgbor You might find this paper interesting: „Five myths about variable selection.” https://t.co/UcszM9EqvP
    view full post

    June 11, 2019

    1

  • Jakub Fronczek, MD, PhD
    @JFronczekMD (Twitter)

    @ValirieAgbor You might find this paper interesting: „Five myths about variable selection.” https://t.co/UcszM9EqvP
    view full post

    June 11, 2019

    5

    1

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @PhDemetri Horseshoe priors work fine, but are not easy to implement. Impossible for med studs. Is VS really needed? See our papers https://t.co/kR1zXMGNiD and https://t.co/R6Dwjt3CSn
    view full post

    May 17, 2019

    1

  • Jakub Fronczek, MD, PhD
    @JFronczekMD (Twitter)

    @laure_wynants @ChristosArgyrop @f2harrell @stephensenn @ADAlthousePhD @Georg__Heinze I use the same paper: „Five myths about variable selection”. It does the trick!
    view full post

    March 29, 2019

    4

  • verbelgde Boris
    @muscovitebob (Twitter)

    RT @laure_wynants: @ChristosArgyrop @f2harrell @stephensenn @ADAlthousePhD @Georg__Heinze has written it down for applied researchers and m…
    view full post

    March 28, 2019

    2

  • Laure Wynants @laurewynants.bsky.social
    @laure_wynants (Twitter)

    @ChristosArgyrop @f2harrell @stephensenn @ADAlthousePhD @Georg__Heinze has written it down for applied researchers and might know of any statements from societies https://t.co/aRg6MNvFCU
    view full post

    March 28, 2019

    18

    2

  • Georg Heinze
    @Georg__Heinze (Twitter)

    @MaartenvSmeden @ADAlthousePhD @htubbscooley_RN @EpiEllie While we're writing on such a piece, consider myths 1 and 2 of https://t.co/R6Dwjt3CSn
    view full post

    February 13, 2019

    2

  • Erik Lenguerrand
    @LenguerrandErik (Twitter)

    RT @MaartenvSmeden: @statsepi These are good references, Darren. I also like what @Georg__Heinze has written about variable selection latel…
    view full post

    January 8, 2019

    6

  • José A. Calvache
    @jacalvache (Twitter)

    RT @MaartenvSmeden: @statsepi These are good references, Darren. I also like what @Georg__Heinze has written about variable selection latel…
    view full post

    January 6, 2019

    6

  • лектор от сатаны
    @rozvidchek (Twitter)

    RT @MaartenvSmeden: @statsepi These are good references, Darren. I also like what @Georg__Heinze has written about variable selection latel…
    view full post

    January 6, 2019

    6

  • Roger Hilfiker
    @RogerHilfiker (Twitter)

    RT @MaartenvSmeden: @statsepi These are good references, Darren. I also like what @Georg__Heinze has written about variable selection latel…
    view full post

    January 6, 2019

    6

  • Maarten van Smeden
    @MaartenvSmeden (Twitter)

    @statsepi These are good references, Darren. I also like what @Georg__Heinze has written about variable selection lately: https://t.co/GULxJFAbyM … and https://t.co/GULxJFAbyM … https://t.co/NFKm2kXY6Z
    view full post

    January 6, 2019

    29

    6

Abstract Synopsis

  • Variable selection in multivariable regression models is popular in transplantation research but can actually complicate analysis by invalidating key statistical tools like P-values and confidence intervals, especially when sample sizes are small or moderate.
  • There are five common myths about variable selection that often lead to its inappropriate use; understanding these misconceptions can help researchers avoid pitfalls.
  • Using expert knowledge instead of relying solely on statistical variable selection methods can often prevent issues, making the analysis more accurate and easier to interpret in transplantation studies.