Roman Folw

[Roman Folw @RomanFolw]

@SashaGusevPosts Tbf, the more accurate interpretation is that the between-group difference is pretty much consistent with null diff, but this is interesting nevertheless.

[Roman Folw @RomanFolw]

@quantitudepod Hello, I'd like to buy one Feyerabend. Wait.

[Roman Folw @RomanFolw]

@paq_research 1. https://t.co/a1Kpowp7XP
2. https://t.co/Qz9i2C4rxp
3. https://t.co/HCa2wOtBxL

[Roman Folw @RomanFolw]

@paq_research Mimochodem, CSI vedle zapojeni do mez. setreni (PISA, TIMSS aj.) a testovani zaku 5. a 9. roc. (vcetne probihajiciho), realizuje i setreni overujici urovne v tematicky zamerenych oblastech. Z poslednich (1) digi. kompetence, (2) ctenarska gram., (3) prirodovedna gram.

[Roman Folw @RomanFolw]

@czstatistika Dobry den, skvele, diky. Lze uz nyni, pripadne bylo by mozne pridat, vypocet podilu z celkoveho poctu osob ve vekove kategorii? Nyni se podily pocitaji pouze z celkoveho poctu dojizdek/vyjizdek.

[Roman Folw @RomanFolw]

@dan_prokop @JanPruha Castecne tam vstupuje i (a) administrativa na strane skol, kazda prihlaska je samostatne spravni rizeni a (b) konani prijimacek na SS (fyzicky limit kapacit).

[Roman Folw @RomanFolw]

@HynekCigler @Fi1618 @sebkova_vladka @MichalSotola @Nataa46640332 @RadkoSablik @JiriStiplCZ @msmtcr 2/2 Samozrejme to samo o sobe neznamena, ze efekt kurzu neexistuje, resp. ze jej neni mozne kvantifikovat. Mne osobne se ale nepodarilo dohledat dostatecne kvalitni vyzkum, ze ktereho bych mohl pri hodnoceni vychazet.

[Roman Folw @RomanFolw]

@HynekCigler @Fi1618 @sebkova_vladka @MichalSotola @Nataa46640332 @RadkoSablik @JiriStiplCZ @msmtcr 1/2 Pred par lety jsem se snazil do tematu proniknout studiem empirickych vyzkumu. Vzdal jsem to uz v reserni fazi. Studie v teto oblasti castokrat neupresnovaly estimand, selhavaly pri definici treatment/control stavu, spolehaly vyhradne na selection-on-observables predpoklad.

[Roman Folw @RomanFolw]

@MeganTStevenson On average, that various theories predict a way larger and long-lasting effects than are observed in reality.

[Roman Folw @RomanFolw]

@LNummenmaa It all comes down to the question whether the LLMs do worse/same/better than humans for this particular task. I could see their point if there is no added-value using them.

[Roman Folw @RomanFolw]

@HynekCigler Pekne, podivam se. Pripomnelo mi to nedavno zverejne setreni mapujici miru (ne)porozumeni formulacim polozek z PHQ.
https://t.co/VyQGmTdppr

[Roman Folw @RomanFolw]

@joefrancis505 Basically same as positivity (overlap). It seems to me that no one discuss it nor care about it in practice.

[Roman Folw @RomanFolw]

@JanKuzebauch Trochu jina informace, ale pro zajimavost. Z setreni PISA 2018 publikovalo OECD vysledky zemi pred i po zohledneni SES zaku. CR takto "ziskala" +10 bodu, mnoho zapodoevropskych zemi "ztratily" (zeme razeny sestupne dle velikosti rozdilu):

[Roman Folw @RomanFolw]

@konvalinka_jan @Vedazije Prvni odstavec je "spravne", pokud by mel popisovat obecny stav veci. Jeho "spravnost" je ale tezko obhajitelna v pripade, ze hodnoti konkretni vystup. V podstate rika: ,,Efekt casopisu A vs. B na kvalitu publikace je minus 1–2 bodu."

[Roman Folw @RomanFolw]

@5_utr Using RCT does not strip you from the necessity of stating those assumptions as well. You just, by design, limit the millions of potential SCM/DAGs into few plausible ones.

[Roman Folw @RomanFolw]

@5_utr What I see as an advantage of CI approach is that you are forced to state all of the assumptions necessary for estimand derivation. Circularity is not a problem since the whole enterprise is build on the premise if A-B-C is true, then the effect of X on Y in population U is true.

[Roman Folw @RomanFolw]

@RexDouglass He should have specified the 'probably'.

[Roman Folw @RomanFolw]

@NobelPrize Perhaps a bit of variance in Y values might help.

[Roman Folw @RomanFolw]

@MishaTeplitskiy Need to see an identification of 'Huge meta-research project' effect on NSF, anyone?