@deirdre_tobias @PlantChompers Interesting that he says (at 10 min.) that FFQs were developed using stepwise regression, which statisticians have been condemning for decades as unreliable. https://t.co/3sNV8mrIXe https://t.co/kB06iYJ2c0
RT @mzloteanu: #statstab #8 Step away from stepwise Thoughts: Pretty straightforward, don't use stepwise regression. It is atheoretical an…
RT @mzloteanu: #statstab #8 Step away from stepwise Thoughts: Pretty straightforward, don't use stepwise regression. It is atheoretical an…
#statstab #8 Step away from stepwise Thoughts: Pretty straightforward, don't use stepwise regression. It is atheoretical and can produce causal inference mistakes; it is also a misuse of p-values. #rstats #pvalues #regression #nhst https://t.co/dPf1i5c
Step away from stepwise https://t.co/0EPpV4RYKm https://t.co/eM2xbCgWtk
@brain_boo @mattansb This is a good overview: https://t.co/F9cGoQvnZv. It’s theoretically unsound in academic settings and methodologically problematic in all contexts.
@mr_shaman_phd @LTF_01 I don't think so. We're missing lots of other things, as @mr_shaman_phd says. It's the same old sh*t (=stepwise) throughout, it's just that high-impact/high-citation publications get more scrutiny. Related: https://t.co/BIppV65VoI
RT @TheodoreLytras: Point of order: please don't do "stepwise regressions"!! (I have a special sensitivity to that topic... https://t.co/o4…
Point of order: please don't do "stepwise regressions"!! (I have a special sensitivity to that topic... https://t.co/o4JBlhK0bY) Here's just a few reasons why you shouldn't: https://t.co/BIppV65VoI
@aNs0gl Je comprends, ms je pense que le souci de parcimonie (rasoir d’Ockam) permet de justifier le choix de stepwise régression (discutable, ms pas rejetable?), avec ttes les pincettes que l’on devra prendre ☺️. Cela dépend évidemment des objectifs de re
@HelenFry1 This isn’t *exactly* it, but pretty close: https://t.co/BZxYd2pt3V
2. Step away from stepwise https://t.co/OC8MR2ejaC
RT @jon_y_huang: @Arrianna_Planey So many ways to tackle this, depends on your context! Classic Biostat: https://t.co/IVtJJmNeBc Modern…
RT @jon_y_huang: @Arrianna_Planey So many ways to tackle this, depends on your context! Classic Biostat: https://t.co/IVtJJmNeBc Modern…
RT @jon_y_huang: @Arrianna_Planey So many ways to tackle this, depends on your context! Classic Biostat: https://t.co/IVtJJmNeBc Modern…
@Arrianna_Planey So many ways to tackle this, depends on your context! Classic Biostat: https://t.co/IVtJJmNeBc Modern Data Science-y: https://t.co/ZC4bvonDVS A synthesis of thought: https://t.co/ecsTSRjayZ
@JSuvilehto @aidangcw @AmandaKMontoya @MarcusCrede I remember reading this at some point but I wonder if the OP has better resources https://t.co/KwP4MlX5e5 For me, learning about DAGs is what made stepwise regression irrelevant & made variable select
@JonLaake Stepwise selection of predictors is problematic. https://t.co/Qpe0UBK4Go
RT @brendannathanl1: Not specific to this article, but applies to a lot of medical research: the ROC has major limitations and should be on…
Not specific to this article, but applies to a lot of medical research: the ROC has major limitations and should be one metric reported among many for model performance. Also, why r we still using stepwise regression?? https://t.co/FBBkwHSgzA
RT @cancerphysicist: Stepwise regression is similarly problematic https://t.co/GEuRuol5md https://t.co/aOmLQBIWyl
Reasons to avoid stepwise regression. https://t.co/KqJ434e6hB
Stepwise regression is a faulty method to select variables. Still prior knowledge is fundamental to select the right model https://t.co/A47CMYwQuA
Stepwise regression is similarly problematic https://t.co/GEuRuol5md https://t.co/aOmLQBIWyl
@Chr_Koenig @Alnajar_MD @StatsForBios try this one: https://t.co/miXWV9sYhT
RT @AugustusPendle1: I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument…
RT @AugustusPendle1: I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument…
RT @AugustusPendle1: I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument…
RT @AugustusPendle1: I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument…
RT @AugustusPendle1: I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument…
I had heard the hype that stepwise regression was dangerous, but this was such an approachable and persuasive argument for its faults! Any recommendations for responses/areas when stepwise is appropriate? #stats #rstats #Bioinformatics https://t.co/YhNFCn3
RT @ChelseaParlett: @VLucet oh gosh there are so many that cover different facets of the issues, here's one i like purely based on the titl…
@VLucet oh gosh there are so many that cover different facets of the issues, here's one i like purely based on the title haha https://t.co/s32O4EtGf7
@POhukainen @epi_twit #dagitty and manual. About DAGs: https://t.co/8d4GaAZD8A https://t.co/zZsqn0Qnaf About stepwise: https://t.co/Blczr3HHgy https://t.co/7O49cdkRND https://t.co/7oOudBiIQM And #dental specific about tbl-2 fallacy: https://t.co/s1tQVpu9
@TenanATC This may or may not be useful: https://t.co/mPg9Dglq7g Found via https://t.co/1lrDO2nVvC
📖
@leonpalafox It has a few issues, but basically it misses important variables, includes nuisance ones, and messes up p-values and CIs (which can be corrected, see: https://t.co/XnwaMCbx8r but often aren’t) . For example: ✅https://t.co/s32O4EtGf7 ✅https
https://t.co/KxnCVBeF51 If you have lots of variables in your model, You are going to get a lot of good, but meaningless fits. Not rocket science, but needs repeating.
RT @ChelseaParlett: Before you @ me, A Stepwise paper: https://t.co/s32O4EtGf7
RT @ChelseaParlett: Before you @ me, A Stepwise paper: https://t.co/s32O4EtGf7
Before you @ me, A Stepwise paper: https://t.co/s32O4EtGf7
@PhDemetri @Kyran_Adams 🧅🦘 🍷🥚🦙🦙 🥗🧊👃🏽🥗🥚 🧶🧅🦄 🍎🥗🦘🥚🦕... https://t.co/s32O4EtGf7
@NikaSeblova @epi_twit @ProfMattFox @EpiEllie https://t.co/G8Ok97s7QY https://t.co/IhnJP2Bx4s Related: bivariate screening https://t.co/3tDr2FYKMI https://t.co/Si8Iz87f5s And finally, source of these and other wonderful sources is @ADAlthousePhD wiki:
We are leery of this use of AIC, which to my reading is designed for testing of a small number of a priori specified models (non-nested). To us this appeared to a use of AIC for stepwise models (for a primer of stepwise regression for model selection see:
In the meantime I found this relatively helpful article: https://t.co/Tk9uEdb76X
@PedsID4Life @ChelseaParlett is someone who can help with cheaper stats, and this paper can help the marmot and all their friends (like frogs!) step away from stepwise: https://t.co/miXWV9sYhT
@StaerkChristian @ChristosArgyrop @f2harrell @stephensenn @ADAlthousePhD step away from stepwise https://t.co/miXWV9sYhT
@FustoloSusanna maybe not the best single references, but these papers titles are hard to beat: “Step Away From Stepwise” https://t.co/4tWASrH28O and “Why Won’t Stepwise Methods Die?” https://t.co/VjTbx4PoLS
RT @adamkvonende: @ADAlthousePhD @ChristosArgyrop @timtriche @f2harrell @stephensenn ?? This is just the most recent paper: https://t.co/Kw…
@ADAlthousePhD @ChristosArgyrop @timtriche @f2harrell @stephensenn ?? This is just the most recent paper: https://t.co/Kw1gbCjL5c