Implied Forced Regression
Implied Forced Regression
#Regression als Bewahrer des Status Quos oder gar als Fortschritt verkaufen. Läuft.
How we tracked down a Go 1.24 memory regression
https://www.datadoghq.com/blog/engineering/go-memory-regression/
One thing that really pisses me off personally is the #regression in terms of #Messenger #Apps.
My personal distaste and dislike for #proprietary, #SingleVendor & #SingleProvider #services like #Signal ¹, #Telegram, #Discord ², #WhatsApp ³, #Slack, #MicrosoftTeams, etc. aside:
WHY is there no #CrossProvider #Messenger to handle that shite?
WHY does everyone of these shitty providers think people want to download their #bloated #WebApp that takes up triple digit Megabytes if not entire Gigabytes and will gobble up all the #RAM and #CPU each of them can??
This problem ain't new and already got solved for corporate social media ages ago! (Not to mention actually good messengers!)
I mean API 0
- style access because obviously none of the platforms will allow, endorse or support such an endeavour and actively fight the developers and users !
So yeah, consider this a call for a @gajim / #Gajim or @pidgin / #Pidgin for garbage platforms!
#statstab #387 Give Your Hypotheses Space!
Thoughts: "It’s tempting to throw a bunch of variables...into a model
...but proceed at your own caution!"
#Mbias #causalinference #collider #moderator #confounder #regression #r #DAG
https://brian-lookabaugh.github.io/website-brianlookabaugh/blog/2025/mutual-adjustment/
Because physical SIMs, like any "cryptographic chipcard" (i.e. @nitrokey ) did all that fancy public/private crypto on silicon and unless that was compromizeable (which AFAICT always necessistated physical access to the #SIM, espechally in pre-#OMAPI devices) the SIM wasn't 'cloneable' and the weakest link always had been the #MNO /.#MVNO issueing (may it be through #SocialHacking employees into #SimSwapping or LEAs showng up with a warrant and demanding "#LawfulInterception"):
Add to that the regression in flexibility:
Unlike a #SimCard which was designed as a vendor-independent, #MultiVendor, #MultiProvider, device agnostic unit to facilitate the the #authentification and #encryption in #GSM (and successor standards), #eSIMs act to restrict #DeviceFreedom and #ConsumerChoice, which with shit like #KYC per #IMEI (i.e. #Turkey demands it after 90 days of roaming per year) und #lMEI-based #Allowlisting (see #Australia's shitty #VoLTE + #2G & #3G shutdown!) are just acts to clamp down on #privacy and #security.
"[…] [Technologies] must always be evaluated for their ability to oppress. […]
And now you know why I consider a #smartphone with eSIM instead of two SIM slots not as a real #DualSIM device because it restricts my ability to freely move devices.
Thus #Impersonation attacks in GSM-based networks are easier than ever before which in the age of more skilled than ever #Cybercriminals and #Cyberterrorists (i.e. #NSA & #Roskomnadnozr) puts espechally the average #TechIlliterate User at risk.
And then think about those who don't have privilegued access to protection by their government, but rather "privilegued access" to prosecution by the state because their very existance is criminalized...
The only advantage eSIMs broight in contrast is 'logistical' convenience because it's mostly a #QRcode and that's just a way to avoid typos on a cryptic #LocalProfileAgent link.
@grmlproject As an old #grml user, today I wanted to use grml 2025.05 for the first time.
I wanted to apt update because I needed an additional package, but that failed with:
E: The repository 'http://snapshot.debian.org/archive/debian/20250514 testing-security Release' does not have a Release file.
Tried again with 2024.12, no problem, worked like a charm like I'm used to it.
A #regression ? Or a surprise from #debian?
Thanks for the great distro btw!
#statstab #385 Understanding and Implementing Robust Regression in R
Thoughts: Not a thorough guide, but shows how simple it is to use a robust regression (rlm vs lm).
#R #regression #robust #statistics #guide #tutorial #outliers #Mestimator
https://www.spsanderson.com/steveondata/posts/2023-11-28/index.html
New Kitten Release
• Fixed regression: Since we switched the h() render function to return a generator (for seamless async support in html templates), two other methods that were calling h() directly – `kitten.safelyAddHtml()` and `kitten.js()` – had begun to fail. This was also affecting the fetchiverse, streamiverse, and kitten-chat family of examples.
Enjoy!
(Run `kitten update` to immediately install the latest version on your development machines. Your production machines will automatically update in the next few hours.)
#statstab #381 In psychology everything mediates everything
Thoughts: Another great explanation of the pitfalls of mediation analysis. Mediation is not a simple procedure and require strong theory.
#mediation #QRPs #correlation #regression #PROCESS
https://steamtraen.blogspot.com/2020/04/in-psychology-everything-mediates.html?m=1
The party told you to stuff ivermectin in your ears and your eyes to avoid the devil's needles.
#statstab #380 Variable Importance Plots {vip}
Thoughts: Not sure how useful these are for researchers, but it's one way to plot the "importance" of your model.
#MachineLearning #lasso #ML #regression #vip #variableselection #r
2024 Time Warp Post:
Left: Staying cool in the Summer heat.
Right: Care Bear training starts strictly from the beginning, so new recruits need to unlearn what they already know.
Art by (left) snugzthegator & (right) @Razz_Beary
@leanlearnlead see, this kind of feature regression on top of being more expensive is why #ElectricCars suck in terms of #UX!
The fact that on top of being slower to load than gas up it's less convenient makes it a no-go for me as I'm not privileged to own a house where I can get a #Wallbox installed!
@haekelschwein Solange #digital unbequemer, teurer (#Preisparität oder gar #Abozwang = teurer!) und #Medien immer weniger #Journalismus abliefern geb' ich mir das nicht!
#TIL at work: sometimes, it may be okay to force a linear regression through the origin.
Normally speaking, you should use "y=ax+b" as model for your linear regressions and correlation indices. Because fixing "b=0" introduces a bias in the estimator "a".
However, today, someone argued that if you're using a regression to measure the predictive power of a model with respect to measurement data, you should fix "b=0".
Now, intuitively this makes some sense, but I haven't been able to find clear proof pro or against it.
So, if you know more about this *and* have relevant references to papers to backup your claim, I'd very much love to hear from you. Statistics is not really my field.
#correlation #regression #stats #statistics #math #maths
Group YCH Ad, Forced Regression
ABDL comic ad, information about forced babying, forced regression
If you want to use the glmnet() in #R for regression or classification but need your data in shape, then use my healthyR.ai package to do it!
Function Reference: https://www.spsanderson.com/healthyR.ai/reference/hai_glmnet_data_prepper.html
Approximately 300 years ago England was deporting subjects for stealing a loaf of bread. Today, the starving are shot for collecting a bag of flour from an aid station. #Regression