Adrian Soto-Mota Profile picture
Apr 15, 2021 15 tweets 6 min read Read on X
Ya está disponible en YouTube la sesión 9 del curso de Metodología y Estadística de la @uiem2017

Hablamos sobre correlación y, sobre todo, de causalidad.

(Link al final del 🧵) Image
Una correlación puede ser perfecta y de todos modos ser inútil.

Por ejemplo, un reloj con la hora de 🇪🇸 tendría una correlación perfecta con la hora de 🇲🇽 pero me haría llegar temprano a mis citas.
Por otro lado, mucha gente cree que, aunque las correlaciones no sirvan para detectar causalidad, por lo menos sirven para detectar:

1) X influye en Y (o al revés).
2) La misma causa influye tanto en X como Y.
3) Una hipótesis valiosa. Image
Pero esto es falso.

El número de películas con Nicolas Cage no influye en cuánta gente se ahoga. El número de ahogados no influye en cuántas películas hay con Nicolas Cage y no es probable que el mismo fenómeno cause ambas cosas.

Tampoco hay hipótesis útiles que buscar aquí. Image
El análisis de causalidad no siempre es intuitivo. Image
Sobre todo en Medicina en donde una exposición puede no ser "causa necesaria", ni "causa suficiente" y de todos modos definitivamente causar enfermedades. ImageImage
Es frecuente encontrar médicos que piensan que los ensayos clínicos aleatorizados sirven para encontrar relaciones causales.

Pero esto también es un error.

(Cambien las agujas de acupuntura de la imagen por cualquier medicamento de su elección). Image
O personas que creen que la "pirámide de la evidencia" tiene algo que ver con la causalidad.

Pero esto también es falso. Image
Hablando de @yudapearl , sería un error hablar de causalidad sin recomendar su libro. Image
En donde, por cierto, nos recuerda que ni Bradford Hill creía que los criterios que propuso son una buena herramienta para encontrar relaciones causales. Image
La mejor forma de evaluar causalidad es con análisis contractuales. Image
Hablamos sobre ellos en la sesión y, si quieren profundizar en el tema, les recomiendo este post de @fhuszar .

En donde los explica con emojis.
inference.vc/causal-inferen…
También hay parábolas sobre un pavo que sabe estadística pero no causalidad y sobre un universo paralelo en donde los gallos son deidades. Image
Espero que les sea útil y que les parezca interesante.

¿Las clases de estadística causan que las personas entiendan sobre correlación y causalidad? 🤔🤔🤔 Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Adrian Soto-Mota

Adrian Soto-Mota Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @AdrianSotoMota

Apr 8
Dear all,

Apologies if I don't engage every time I am tagged. This month, I'm overseeing hospital care, and I'm very busy.

Thank you all for your interest in our work! Your support keeps us motivated and your questions & criticism keep us improving.

I'll address them here 🧵
Some have pointed out that SuppTable1 shows that people with ⬆️CAC (>100) had 2.4 annual PAV change. Is this too fast?

Sure! Many have defined fast plaque progression as a yearly change >1% (i.e. PARADIGM reg).

Is the keto-induced ⬆️ApoB to blame for their plaque progression?
We tested (many times) if ⬆️ApoB was a significant plaque progression predictor when someone had ⬆️baseline plaque.

This model looks for an interaction between BLplaque and ApoB. In other words, if ApoB is predictive when BL plaque has different values. Image
Read 15 tweets
Apr 7
⚠️📣NEW PAPER📣⚠️
@JACCJournals
with @nicknorwitz, @realDaveFeldman,
@DrRagnar, Venkat Manubolu, April Kinninger, @jayearls &
@BudoffMd

(URL🔗 at the end of this 🧵)

Longitudinal CCTA-LMHR data is HERE!!!

And...

YES!!!
It's awesome!!!

But, first...
Those who know me, know, that the hill I'll die defending is "Mount Nuance."

I'm an internist. I'm physiologically incapable of making straightforward statements.

Don't take my word, ask @DGlaucomflecken about my kind.

If you expect to hear "X is Good" & "Y is Bad," well...
We've been accused of "spreading confusion." We aren't. We're simply showing that the ApoB story isn't as simple as we thought.

No, we didn't find LMHRs don't have plaque progression.

No, we didn't find anyone should stop their meds if they "Go Keto".
Read 24 tweets
Jan 17, 2024
⚠️📣NEW PAPER🔔‼️
@AJCNutrition
(link at the end👇🏼🧵)

With Yusceli Flores, @nicknorwitz, @realDaveFeldman, Mark Pereira, Goodarz Danaei, and @davidludwigmd

A meta-analysis (MAA) + individual participant data (IPD) analysis of 41 RCTs entailing the #LEM and the #LMHR phenotype. Image
In 2022, we published an observational study suggesting BMI was a mayor player in LDL-C changes during carbohydrate restriction:

However, we didn't have dietary intake reports, all data was self-reported, and all participants self-selected their diet. doi.org/10.1093/cdn/nz…

Image
Image
Thus, we sought RCTs that documented dietary intake and randomized people across the BMI spectrum to physiologically relevant carbohydrate restriction.

We found 41 elegible RCTs.

Additionally, we compared the influence of saturated fat intake with BMI's on LDL-C changes. Image
Read 18 tweets
Mar 7, 2023
What do the liver, the thyroid, 🩸phosphorus & your 🩸glucose have to do with migraine?

⚠️🚨NEW PUBLICATION 🚨⚠️
@SciReports with @DrElenaGross

nature.com/articles/s4159…

Many don't think about migraine (or most chronic 🧠conditions) as metabolic diseases.

But they are! 🧵
The human 🧠 is the most amazing thing around. Intuitively, this amazingness does not come cheap!!!

The human 🧠 is exceptionally expensive to sustain energetically.

Compared to other primates, we have the most expensive 🧠👇🏼

doi.org/10.1038/nature…
I think it's reasonable to think that, if anything is wrong with someone's energy management system (a.k.a. their Metabolism), one of the first tissues where thing could go wrong, it's the most energy demanding of them 🧠.
Read 15 tweets
Mar 6, 2023
2nd erythritol 🧵.
(Shorter than the 1st one I✋️)

Given that some of the pieces of evidence are "similar" and some of the researchers are the same?

Is erythritol the new TMAO?

Why I'd advise avoiding erythritol but not choline (TMAO)?

First 🧵
threadreaderapp.com/thread/1630971…
TMAO = Trimethylamine N-oxide.

After we eat choline-rich foods, SOME gut🦠 produce TMA (Trimethylamine) which is then transported to the liver and converted into TMAO.
Now, in 2016, some of the same researchers in THE erythritol paper published a set of studies linking TMAO with heart disease.

cell.com/fulltext/S0092…
Read 9 tweets
Mar 1, 2023
Not mutually exclusive, but most people said they'd prefer seeing rather than joining a discussion to analyze THAT erythritol paper, and I have some airport hours ahead...

Here go🧵.

In case you've been living under a rock, this is the paper:

nature.com/articles/s4159… Image
Relevant disclosures for interpreting my interpretation?

I'll speak more as a clinician clinician who sees many ⬆️CVDrisk patients following a low-carb diet than as a healthy adult who follows a low-carb diet and occasionally consumes erythritol.
Things I liked about the paper = ✅️.

Things I didn't like (and matter) = ⚠️

Things I didn't like (and are nitpicking) = 🧐.

🩸[] = blood concentration.

CVD = CardioVascular Disease.

MACE = Major Adverse Cardiovascular Events.

Why we "liked" erythritol in the first place?
Read 32 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(