Andy Siddall Ph.D.

I learn a huge amount from papers that are adjacent to, or completely outside, my research areas that I will probably never have a reason to cite. A friendly reminder that paper impact will be (immeasurably) larger than just # of citations! #AcademicTwitter #AcademicChatter

The 4th International Physical Employment Standards Conference (IPES) 2023 being held Bond University with the Tactical Research Unit is here: See below for registrations, abstract submissions, and sponsorship opportunities. bond.edu.au/researchers/re…

The 4th International Physical Employment Standards Conference (IPES) 2023 being held Bond University with the Tactical Research Unit is here: See below for registrations, abstract submissions, and sponsorship opportunities. lnkd.in/g5KwnxT3

Kind reminder that "validated" generally means "someone looked at it" not "someone looked at it and judged it to be good and useful"

#AddAStatisticalWordorPhraseRuinAMovie In the spirit of Jimmy Fallon, add a statistical word or phrase to a movie title and change its meaning. I'm putting together a top ten list of these for the @TheASAPodcast and would love to add yours to the list. Here are some examples:

@tessamaroni Presenting our research on Entry Fitness And Subsequent Physical Performance Change In Recruits Across British Army Basic Training Courses. [Board No. 178] this morning at #ACSM2022 in San Diego. @chiuni @OPRG_UniChi

With COVID, and all infections: Symptoms occur for at least 2 VERY different reasons 1) Immunity Fighting the infection (fever, congestion) 2) The infection winning & causing harm (lose smell, breathing issues) Symptoms do NOT define if you are infectious, the virus does! 1/

PhD thesis submitted today by @onthegrape_vine !

@mattansb @statsepi @MaartenvSmeden I don’t know why but I read this like the Pokémon game meme: Covariate used Noise. It had no effect.

@statsepi @MaartenvSmeden We measured noise, and it has no effect - wow!

Yeah, that is not how it works

Fantastic work and congratulations to @vickyedwards14 for pulling together this tough dual submission from @OPRG_UniChi monitoring officer training! Fun figure: disparity in CHO/PRO intake timing between when training could allow core meals (a, d) vs. field exercise (b, e)

Pleased to have these two published in @IJSNEMJournal 🥳 describing energy balance & availability, and the energy & macro distribution in Officer Cadets at @RMASandhurst over their 44 week commissioning course. doi.org/10.1123/ijsnem… doi.org/10.1123/ijsnem… @OPRG_UniChi

@TenanATC @ExPhysStudent Thanks- This is precisely the sort of thing that prompted my question. Seeing one-to-one time-aligned data from an individual, and wanting to know if/what approaches exist to statistically do the “visual checking”. Like “pattern/shape” agreement - if that makes sense!

@ExPhysStudent @JumpXav I've typically focused on individual 1-to-1 data series which are time-aligned and then visually inspecting their "agreement" with whatever bounds are defined across that time series.

Will be completely out of my depth- but what approaches exist for equivalence/agreement across high frequency time-series data? (E.g time-aligned HRV in 2 devices). Many claim “correlation” by superimposing one longitudinal pattern on top of another. Ideas? (Perhaps @TenanATC?)

@ExPhysStudent @TenanATC Perfect, thanks - that makes sense

@JumpXav @TenanATC Yeah, my suggestion would be to use more than 1 analysis. The answers from cccrm and SimplyAgree could differ slightly (good to investigate "why" they may differ in conclusions). Matt's emmeans suggestion would be contained within the "bias" reported in the loa_mixed results.

@ExPhysStudent @TenanATC Aaron, thanks for saying this - I would be interested whether there are some who would not define this as agreement - or it’s close enough (not wordplay intended). I suppose it may also depend on what each epochs of time you repeatedly apply this over.

@TenanATC @JumpXav Hmm, but wouldn't that only be testing if there are differences in central tendency (mean response) and not "agreement"?

@TenanATC @ExPhysStudent I think this was one of my concerns - was whether average differences over time could be used for “sameness” - the methods above are helpful. I will look forward to trying these all out on day!

@ExPhysStudent @JumpXav But he seemed to ask for both an 'agreement' and 'equivalence' type analyses. They're different, but I think, if you make it extensible across time (not in that above code), it gives an idea as to "how the same" the series are at different time points.

@ExPhysStudent @TenanATC Glad you’ve joined the party too Aaron! Aware that agreement and equivalence are different, as LoA and equiv testing are sometimes used in tandem to show different concepts, so this was an either/or type question to see what methods existed!

@TenanATC @JumpXav Yeah, I get what you are saying, but in building the SimplyAgree package I found that some people to use two one-sided tests of equivalence (TOST) to refer to limits of agreement as well doi.org/10.1080/194663…

@TenanATC Thanks Matt - I figured you might have! I am not (currently) looking to run an analysis, but I will have use in the near future. But have just been seeing a lot of claims without an analysis of this type- so it made me wonder what existed and/or is actually appropriate.

@JumpXav Yeah, so I've actually done this in some consulting work. Are both devices on the same scale? That makes life a lot easier. If they are, you just need to define your equivalence bounds. Build your model with a time effect & device effect, then run the equivalence test as desired.

It remains surprising to me that a limited analysis, that told us little we didn't already know, not only made the front page of the NYT, but seems to have been taken as "case-closed" on non-lab origins of sars-2.

There is no scientific merit to the use of a simplified heatmap (kernel density estimate) to represent residential addresses of early COVID cases in Worobey et al (2022) zenodo.org/record/6299600. @MichaelWorobey @acritschristoph @K_G_Andersen @stuartjdneil @edwardcholmes @arambaut

JUST IN: a new DNA sequence, known as TLDR, has been identified. It is so long that the polymerase becomes bored and stops transcribing it.