bbolker

@bbolker@fediscience.org

Ecology, evolution, epidemiology, statistics at McMaster University.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

bbolker, to random

(sigh) me, a 20-year user of R, web-searching for "debugging R S4 methods" for the umpteenth time ... #rstats

bbolker, to random

Has anyone else run into problems using the patchwork package with the new ggplot2 version 3.5.0 ... ?? (Probably reverting to 3.4.4 for now but wondering if anyone else is getting bitten or if it's just me - I don't see anything on https://github.com/thomasp85/patchwork/issues, but https://github.com/easystats/performance/issues/691#issue-2164468933 ...

eliocamp, to random
@eliocamp@mastodon.social avatar

Woah, this new ggplot2 release is a banger!

https://www.tidyverse.org/blog/2024/02/ggplot2-3-5-0/

#RStats

bbolker,

@eliocamp "Your programmers were so preoccupied with whether or not they could ..." https://knowyourmeme.com/memes/your-scientists-were-so-preoccupied-with-whether-or-not-they-could-they-didnt-stop-to-think-if-they-should

bbolker, to random

I want to tell my students to convert human-friendly column names ("Weekly Rate") into computer-friendly names (weekly_rate), but think they should also maintain the human-friendly names as attributes. I know there's Hmisc::label(); does anyone have other suggestions for R packages/frameworks supporting this kind of functionality? (I've also occasionally found that Hmisc::labels() attributes will cause odd errors, although of course I don't have a reprex right now ...)

davidbraze, to random
@davidbraze@mstdn.social avatar

within the last year or so @Posit has taken a big tranche of venture capital. VC wants one thing: a massive return, 10x minimum. there are two ways to get that. first is to take a company public. I don't think that's likely here. second is a buy out by a major player like Oracle, Microsoft, Amazon, or another of that ilk. I guess this is what posit's sugar daddy is likely angling for. this isn't unprecedented in the #rstats world. who else remembers Revolution R?

https://mastodon.social/@awong234/111751453549394950

bbolker,

@benschneider @davidbraze @Posit I'd be interested in a citation too.

AFAIK they're still a public benefit corporation; the latest annual report I can find is from 2021
https://posit.co/about/pbc-report/ "To fulfill its beneficial purposes, RStudio intends to remain an independent company over the long term."

The RStudio-to-Posit FAQ states that nothing will change: https://posit.co/blog/rstudio-is-now-posit/

The latest B corporation report is from 2022 (I think):

https://www.bcorporation.net/en-us/find-a-b-corp/company/rstudio/

bbolker,

@benschneider @davidbraze @Posit I am curious, though, if anyone can help locate later reports. 2021 annual report: https://posit.co/about/pbc-report/ says "As a PBC, RStudio publishes an annual report that describes the public benefit we have created, along with how we seek to provide public benefits in the future. This is the third of these reports." Presumably at least a 2022 report should be available somewhere, even if the 2023 report isn't out yet since it's still early in Jan 2024?

bbolker, to random

Just read "by comparing the output with SAS®, which is the current gold standard for statistical programming" in the abstract of a paper published in 2020 ... (Granted SAS is really good, but "gold standard"?)

UchidaMizuki, to rstats

@rstats

R is one of the best languages for data science.

But what about optimization? In optimization, Julia is attractive to me because it allows for automatic differentiation.

Can we do automatic differentiation in R?
(re-draft)

bbolker,

@UchidaMizuki @rstats I agree that this is a place where Julia shines. There have been a variety of attempts (https://www.google.com/search?q=autodiff+R) but all are either rudimentary (in fact base-R deriv() is essentially doing autodiff for a v. limited subset of the language) or involve a back-end in another language (Python, C++ via TMB https://github.com/kaskr/RTMB or Stan https://www.jchau.org/2022/01/24/automatic-differentiation-in-r-with-stan-math/, Julia ...) RTMB may be the closest at this point.

Cmastication, to random
@Cmastication@mastodon.social avatar

One of my frustrations with the mandatory IT security training we're all forced to take at work is that most of these punish us for being risk averse. I would not open any of the emails they show me. But I get dinged for that. But my approach is less likely to result in malware infestation.

Of course, given my druthers, I would also disassemble my computer, put the bits in a box, light that box on fire, and walk into the woods in a dramatic scene backlit by the burning box of computer bits.

bbolker,

@1010is10 @Cmastication what were the general criteria for when you should delete vs report? (My own criteria would be "is this pretty obvious/typical or is it dangerously sophisticated/a new type?" and "do I think it's probably been reported by someone else already?")

gaborcsardi, to random
@gaborcsardi@fosstodon.org avatar

The new evercran project (https://github.com/r-hub/evercran#readme) helps you run old (or ancient) versions of R, together with old (or ancient) snapshots of CRAN.

bbolker,

@gaborcsardi I kind of want to try this out with the 0.64.x range, the first versions of R I used ... #oldbrag #rstats

ojala, to random
@ojala@mastodon.nz avatar

deleted_by_author

    •
    bbolker,

    @ojala I don't disagree at all. The ever-present question: how do we get more resources/attract more people to work on the R mixed-model infrastructure? I can think of dozens of relatively easy/low-hanging-fruit ways to make things better, and a few medium-challenge ones, but software development is not really my day job (teaching, mentoring grad students, administration, eco/evo/stats research that is not 'just' technology transfer ...)

    bbolker,

    (1/?) @ojala I have a lot to say about this (surprise). (1) It's true that R favours a chaotic ecosystem (among other things, there's an incentive for academics to work on their own packages rather than join a group effort) - as another commercial comparison:I don't know who works on SAS MM these days, but AFAIK Stroup and Schabenberger were responsible for a huge amount of PROC GENMOD/NLMIXED etc.. ...

    bbolker,

    @ojala (2/?) Even if you decide on One Package To Rule Them All, development is too slow. lme4 really wants refactoring. glmmTMB is newer and easier to develop, but there are still ZERO people who are paid to develop and/or maintain either of these packages. (I honestly don't know where Paul Bürkner finds the time to develop brms, which is one of the most full-featured packages around ...) Doug Bates (emeritus) works on MixedModels.jl in Julia.

    bbolker,

    @ojala (3/?) Probably the biggest bang for the buck is to invest effort in 'umbrella' packages (broom.mixed, multilevelmod, etc.). (4) I could/should invest effort in hustling for resources rather than complaining, but honestly that's a lot of work too, and much less fun than doing actual development (neither of which I get any official credit for ...)

    bbolker,

    @ojala No (which IMO is a good thing); merit pay evaluation is more holistic than that, and criteria are set separately by each department (I'm already at the top academic rank, unless I get nominated for some 'distinguished scholar' thing)

    Private
    bbolker,

    @DrSimonCarr @tiago @academicchatter Not sure that Google is morally superior to Clarivate https://en.wikipedia.org/wiki/Clarivate ("[~50% profit margin ...] in the range of entertainment software and tobacco industries, and ahead of pharmaceutical, oil/gas and cell phone industries ... anti-competitive practices in a highly oligopolistic academic database market ...") Agree that #WebofScience is a way better search interface for academic articles ... #Scopus may be even better but ugh, Elsevier!

    bbolker,

    @rspfau @DrSimonCarr @tiago @academicchatter What discipline? (PubMed is also good if it indexes everything you need, or mathscinet for math ...)

    bbolker,

    @rspfau @DrSimonCarr @tiago @academicchatter Hmmm, is WoS really so bad for pop gen? Surprising ...

    bbolker,

    @danriggins @DrSimonCarr @tiago @academicchatter @OpenAlex Hmm. " If you want to explore the data as a human, you may be more interested in OpenAlex Web. This web interface is currently in the alpha stage of development, with a beta launch coming soon " ... that puts me off a little bit - may try it out the next time I have something programmatic/bibliometric to do, but I'm not sure I want to learn a new API to do a quick search ... ? Or am I missing something?

    Cmastication, to random
    @Cmastication@mastodon.social avatar

    I’m sure I’ve read before that “floating point math is not associative” and my eyes glazed over. And I know that floating point math IS deterministic… i.e. same numbers in = same result every time.

    I spent a week understanding why a Spark job was not deterministic. The crux is that in Spark, values are spread across workers in a non deterministic way. So then the order of numbers fed into math operations varies from one run to next. Which makes output non-deterministic.

    bbolker,

    @Cmastication @johndcook Fun r-help thread about non-deterministic results on a non-distributed platform due to auto-parallelization in Intel's MKL compiler (including an ignorant comment from me): https://stat.ethz.ch/pipermail/r-help/2023-August/477904.html See here https://www.intel.com/content/www/us/en/developer/articles/technical/introduction-to-the-conditional-numerical-reproducibility-cnr.html for instructions on how to overcome this (set MKL_CBWR=COMPATIBLE)

    bbolker, to random

    Anyone know offhand if there is a way to include multiple bibliographies in a single Quarto document (e.g. 'peer-reviewed articles' and 'books', or 'core references' and 'auxiliary refs') ? I would use the bibunits package for this in TeX ... https://ctan.org/pkg/bibunits?lang=en #quarto #bibliography

    KorinnaAllhoff, to random German
    @KorinnaAllhoff@ecoevo.social avatar

    Dear Msc students applying for a ,

    I've recently reviewed almost 48 job applications. Can you guess what the most common mistake was?

    Well. It's ok to come from a different background. It might even be a bonus! But please explain WHY you want to switch fields and HOW your unique skill-set might be helpful to get the given job done. Please tell me why your past achievements are important for THIS job and why you want to work in THIS team.

    Good luck!

    bbolker,

    @KorinnaAllhoff Agree. Even more strongly: "Recognize that graduate training is a two-way street. If I'm going to take you on as a student who is switching fields, what makes it worth my while to help you with the extra training/orientation you'll need? Can you tell me about a few steps you've already taken/time you've invested to start getting yourself up to speed in this new area?" (I am sympathetic to students switching fields, I was one once ...)

    lakens, to random
    @lakens@mastodon.social avatar

    After not donating blood for 8 years (first a medical issue, then I lost the habit) I gave my 49th donation last Monday. Unlike your grandmother they are not judgmental at all and just happy to see you after 8 years. If you are like me, sign back on here: https://www.sanquin.nl/bloed-doneren

    bbolker,

    @lakens Unless you live in Canada and spent 3 years in the UK during the 'mad cow' period (I could donate when I go back to the US, I think, but it's not a big priority when I'm visiting family)

    ojala, to random
    @ojala@mastodon.nz avatar

    deleted_by_author

    •
    bbolker,

    @ojala At a quick skim this looks like it's mostly about circular response variables. In case this hasn't occurred to you already, I think the simplest thing will be to use sin(direction), cos(direction) as predictors ...

    Private
    bbolker,

    @statstas @rstats I've probably posted this before, but here's another (highly entertaining) R-history presentation (from 2012) by Pat Burns: “Inferno-Ish R: Or, How I Learned to Stop Worrying and Love the Bomb.” Presented at the CambR, May 29. https://www.burns-stat.com/documents/presentations/inferno-ish-R/.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • mdbf
  • ngwrru68w68
  • tester
  • magazineikmin
  • thenastyranch
  • rosin
  • khanakhh
  • InstantRegret
  • Youngstown
  • slotface
  • Durango
  • kavyap
  • DreamBathrooms
  • JUstTest
  • tacticalgear
  • osvaldo12
  • normalnudes
  • cubers
  • cisconetworking
  • everett
  • GTA5RPClips
  • ethstaker
  • Leos
  • provamag3
  • anitta
  • modclub
  • lostlight
  • All magazines
    •