Thread by @ncdave4life

@NolanCarlton2 @BenKoby1911 @RushhourP @ChrisBBacon3 @Climatehope2 @Huroner2 @Willard1951 @Devonian1342 @fischerfischy @Mark_A_Lunn @AristotleMrs @Michael_D_Crow @balls95652097 @Anvndarnamn5 @sandcastle1975 @Jaisans @Hji45519156 @Data79504085 @S_D_Mannix @JusticeTrudeau @JimBlack48 @glinch72 @DenisDaly @stitchnstab @CarrudoDon @Veritatem2021 @LiveLifeBK24 @TheDisproof @priscian @Ceist8 @Joeyd87745119 @waxliberty @SuperFoxyLoxy @JaapTitulaer @wjack76995 @Rocky35418823 @NobaconEgbert @JustThi30117912 @SniemN @BointonGiles @ammocrypta @SeekerTheGreat1 @ubique60 @CDMarshall7 @EthonRaptor @RMcgillss @paligap17 @MaggieL @Willy1000 1/4》Nolan, what you did gives the final period a small weight at the beginning but huge weight at the end. So if that final period happens to be below average, you'll generate a downward-blade hockey stick, or if it's above average you'll generate an upward-blade hockey stick.

@NolanCarlton2 @BenKoby1911 @RushhourP @ChrisBBacon3 @Climatehope2 @Huroner2 @Willard1951 @Devonian1342 @fischerfischy @Mark_A_Lunn @AristotleMrs @Michael_D_Crow @balls95652097 @Anvndarnamn5 @sandcastle1975 @Jaisans @Hji45519156 @Data79504085 @S_D_Mannix @JusticeTrudeau @JimBlack48 @glinch72 @DenisDaly @stitchnstab @CarrudoDon @Veritatem2021 @LiveLifeBK24 @TheDisproof @priscian @Ceist8 @Joeyd87745119 @waxliberty @SuperFoxyLoxy @JaapTitulaer @wjack76995 @Rocky35418823 @NobaconEgbert @JustThi30117912 @SniemN @BointonGiles @ammocrypta @SeekerTheGreat1 @ubique60 @CDMarshall7 @EthonRaptor @RMcgillss @paligap17 @MaggieL @Willy1000 2/4》Both of those those plots would be deceptive, because both create the illusion of a change in trend which is not real. You could feed random trendless data to that algorithm, and generate a high percentage of "hockey sticks," all with the blades at the right end.

@NolanCarlton2 @BenKoby1911 @RushhourP @ChrisBBacon3 @Climatehope2 @Huroner2 @Willard1951 @Devonian1342 @fischerfischy @Mark_A_Lunn @AristotleMrs @Michael_D_Crow @balls95652097 @Anvndarnamn5 @sandcastle1975 @Jaisans @Hji45519156 @Data79504085 @S_D_Mannix @JusticeTrudeau @JimBlack48 @glinch72 @DenisDaly @stitchnstab @CarrudoDon @Veritatem2021 @LiveLifeBK24 @TheDisproof @priscian @Ceist8 @Joeyd87745119 @waxliberty @SuperFoxyLoxy @JaapTitulaer @wjack76995 @Rocky35418823 @NobaconEgbert @JustThi30117912 @SniemN @BointonGiles @ammocrypta @SeekerTheGreat1 @ubique60 @CDMarshall7 @EthonRaptor @RMcgillss @paligap17 @MaggieL @Willy1000 3/4》Do you wonder why smoothed plots always end before the end of the data, instead of using shortened analysis periods tp go all the way to the end, like you did? It's because a shortened final analysis period often generates an illusory appearance of deviation from the norm.

@NolanCarlton2 @BenKoby1911 @RushhourP @ChrisBBacon3 @Climatehope2 @Huroner2 @Willard1951 @Devonian1342 @fischerfischy @Mark_A_Lunn @AristotleMrs @Michael_D_Crow @balls95652097 @Anvndarnamn5 @sandcastle1975 @Jaisans @Hji45519156 @Data79504085 @S_D_Mannix @JusticeTrudeau @JimBlack48 @glinch72 @DenisDaly @stitchnstab @CarrudoDon @Veritatem2021 @LiveLifeBK24 @TheDisproof @priscian @Ceist8 @Joeyd87745119 @waxliberty @SuperFoxyLoxy @JaapTitulaer @wjack76995 @Rocky35418823 @NobaconEgbert @JustThi30117912 @SniemN @BointonGiles @ammocrypta @SeekerTheGreat1 @ubique60 @CDMarshall7 @EthonRaptor @RMcgillss @paligap17 @MaggieL @Willy1000 4/4》Any competent statistician could explain to you what's wrong with that approach. (Even Tamino might be able to… maybe.)

A plot in which a single data value (or a particular subset of data values) is given VARYING weight is simply wrong.