[email protected]

Thunderstorm on the way back from Paris

At the heart of it, there is no real difference between what AI researchers do with data and what statisticians do with data. The differentiator is the statistics community is generally unwilling to build things they don't understand.

PARIS QUALIFIED! 🇫🇷 #Ireland7s | @TritonLake

💚

#shouldertoshoulder #GrandSlam

A group representing practitioners, NGBs and third level institutions are working together to develop an Irish Sport and Exercise Science Association. Group are keen to gather views of the community in Ireland via the survey below: surveymonkey.com/r/ISESA

@tanya_shapiro Because I hate tidyverse; for (i in 1:nrow(df)) { if (is.na(df$people_fully_vaccinated[i])) { df$people_fully_vaccinated[i] <- df$people_fully_vaccinated[i-1] } }

@drdeanjmiller And any experience of trying to get raw accelerometer data out of commercial devices, to run it through (e.g.) GGIR? Preferably with a bulk export of even an API, rather than having to ask nicely in an e-mail?

@drdeanjmiller Would it be fair then to suggest that little/no impact on the overall finding of validity for 2-stage agreement, maybe an effect on multi-stage agreement, which was fairly poor anyway?

@drdeanjmiller @DrSianAllen @_Matthew_Shaw Great study, and long overdue comparison to PSG. Any thoughts on the impact (or not) of the first night effect?

@brodriguesco @JosiahParry Respectfully disagree, I find loops easier to write, read, and far far easier to debug than the equivalent function

@JosiahParry Loops aren't bad per se, and allow to solve many problems. But there hard to write, harder to read, even harder to extend (and that's not an R specific thing). Very often, functional programming provides a cleaners and simpler way of expressing looping.

@arthurturrell Used to be berryFunctions for is.error().... then I learned how to use tryCatch 😂

What's your favourite non-obvious package in R? As in, not ggplot or dplyr or any of the really popular ones (can still be tidyverse though). #rstats #datascience #R (Mine is probably fixest)

@Tmonayy @arthurturrell beep(sound = 9) on max volume really annoys the wife

@arthurturrell beepr. It allows you to play sounds. Super useful when I do large data pulls and they take a while. Plays a sound when it's done.

@Dr_DaveN @LightningCycle @GoZwift Was that the Dave Nichols peaks special I remember, with a massive climb straight after the first stop

@rsfrankl @turinginst Ireland close enough?

@AnthonyTeacher Is hashing the passwords using openssl an option? Or do you need the users plaintext password to 'forward' to the database?

#rstats I am building a shiny app that will connect to a database. what would be the best (safest) way to store the database credentials for this app? I tried keyring, but it seems to require interactive input of the password.

@f2harrell @MarkAnnuncio @statsepi Not to mention the irrational (IMO) fear of $ and [ ]

@AndyW_87 @MarkAnnuncio @statsepi Agreed. Pipes take a non-problem and make things worse in the majority of cases. Plus new #rstats users who start off with pipes don't learn rudimentary solutions to problems.

@ERDonnachie @buidiengiau @rstats_tweets Spot on

Still can't stand why we need to add option na.rm=TRUE in R, which I think should be default. Any use case we need na.rm=FALSE? Maybe nope. #rstats

@MarkAnnuncio @f2harrell @statsepi 100% Empty function calls always worry me. Plus I find piped syntax much harder to read

@f2harrell @statsepi I think sometimes pipe can be nice but frequently it can obscure argument logic, especially when it's not common for ppl to label each argument So I do think there is argument to not use pipes for readability. I think that's important especially when sharing code

@eo_performance @DanielKadlec Sadly not surprising from the discipline that invented MBI

"We found that 85% of the 20 most highly cited meta-analyses in strength and conditioning research contained statistical errors." 🤯 @DanielKadlec link.springer.com/article/10.100…