Ben

[Ben @Weidmann]

@Sakiera_Hudson In our lab work we found that men are typically more willing to be leaders, but that this is unrelated to the causal contribution leaders make to teams (in fact, self-promoters tend to perform a bit worse than randomly-chosen leaders) https://t.co/UtJveEUrNv

[Peter Singer @PeterSinger]

The world’s 1st octopus farm would cram a MILLION of these smart and sensitive creatures into tiny tanks before painfully freezing them to death – but a massive public outcry can help stop it. Sign and share now! #BanOctopusFarming https://t.co/sgZws8ZpT3

[Ben @Weidmann]

@_criminologist @Martell_Tom @WhatWorksUK @JohnDeke That said, I think we should prioritise risks based on how bad they seem to be in practice...and on current best evidence (for ed evaluations) I worry less about attrition than I do about other things

[Ben @Weidmann]

@_criminologist @Martell_Tom @WhatWorksUK Good point - in the context of small effect sizes, even small biases are a big deal in % terms. @JohnDeke has a cool paper on this (https://t.co/CPMmS6fzz0)

[Ben @Weidmann]

@Martell_Tom @_criminologist @WhatWorksUK Strongly agree! If anyone's keen on expanding this work, I'd be happy to pitch in

[Ben @Weidmann]

@Martell_Tom Not at the moment, but I think the number of studies in the archive has grown quite a bit...so it would be well worth following up!

[Ben @Weidmann]

@Martell_Tom This is above-average attrition, but I think it's getting too much attention. Research on attrition bias in EEF trials suggests that it might not be a killer issue (https://t.co/BX5OnLZwQY)

[Ben @Weidmann]

@mike_t_sanders This is a great summary! In another set of within-study comparisons, we found more optimistic news about matching. (https://t.co/TgtH50PchF)

[Ben @Weidmann]

@fatih_unlu_R @JoshuaSGoodman @catarm2010 @LMiratrix Researchers / policymakers who are interested in K-12 (and attainment outcomes) often have a big advantage here, as non-randomized studies seem to suffer from smaller selection bias

[Ben @Weidmann]

@fatih_unlu_R @JoshuaSGoodman @catarm2010 @LMiratrix Interesting! Thanks for flagging the paper. Definitely agree that context matters for the trustworthiness of observational studies (and a big part of this is "do we have reliable pretest)

[Ben @Weidmann]

@JoshuaSGoodman @catarm2010 @fatih_unlu_R @LMiratrix The main weakness of our analysis is that "selection into a wait-list RCT" might differ in important ways from "business-as-usual selection into a program/practice". We discuss this in the paper! Overall, I think the analogy is a pretty good one (but not perfect)

[Ben @Weidmann]

@JoshuaSGoodman @catarm2010 @fatih_unlu_R @LMiratrix In our JPAM paper, we explicitly study the effect of 'selection into a program': i.e. we compare schools/pupils who have an interest in a program and 'opt in' (but are randomized to the control arm!) with a group who weren't interested