NaiveTortoise

Comments

TurnTrout's shortform feed

I'm curious what sort of things you're Anki-fying (e.g. a few examples for measure theory).

How can we lobby to get a vaccine distributed faster?

Minor correction: I think you mean Alex Tabarrok (other author on MR).

Probability vs Likelihood

I find it helpful to have more real world examples to anchor on so here's another COVID-related example of what I'm pretty sure is likelihood / probability confusion.

Sensitivity and specificity (terrible terms IMO but common) model and respectively and therefore are likelihoods. If I get a positive test, I likely have COVID, but it still may not be very probable that I have COVID if I live in, e.g. Taiwan, where the base rate of having COVID is very low.

Three Open Problems in Aging

I'm the person starting to work on the senescence-induced senescence problem. Happy to chat more about current thoughts / plan (I am open to trading marginal time for relatively small amounts of $ but also happy to just talk about what I plan to do anyway). Feel free to DM me.

Open & Welcome Thread – November 2020

The first way to treat this in the DAG paradigm that comes to mind is that the "quantitative" question is a question about a causal effect given a hypothesized diagram

On the other hand, the "qualitative" question can be framed in two ways, I think. In the first, the question is about which DAG best describes reality given the choice of different DAGs that represent different sets of species having an effect. But in principle, we could also just construct a larger graph with all possible species as s having arrows pointing to $ X $ and try to infer all the different effects jointly, translating the qualitative question into a quantitative one. (The species that don't effect $ X $ will just have a causal effect of $ 0 $ on $ X $.)

To your point about diversity in the wild, in theoretical causality, our ability to generalize depends on 1) the structure of the DAG and 2) our level of knowledge of the underlying mechanisms. If we only have a blackbox understanding of the graph structure and the size of the average effects (that is, $ P(Y \mid \text{do}(\mathbf{X})) $), then there exist [certain situations](https://ftp.cs.ucla.edu/pub/stat_ser/r372-a.pdf) in which we can "transport" our results from the lab to other situations. If we actually know the underlying mechanisms (the structural causal model equations in causal DAG terminology), then we can potentially apply our results even outside of the situations in which our graph structure and known quantities are "transportable".

Three more stories about causation

Oh I see, yeah this sounds hard. The causal graph wouldn't be a DAG because it's cyclic, in which case there may be something you can do but the "standard" (read: what you'd find in Pearl's Causality) won't help you unless I'm forgetting something.

An apparently real hypothesis that fits this pattern is that people take more risks / do more unhealthy things the more they know healthcare can heal them / keep them alive.

Three more stories about causation

A few minor comments. Regarding I, it's known that the direction of (or lack of) an arrow in generic two-node causal is un-identifiable, although there's some recent work solving this in restricted cases.

Regarding II, if I understand correctly, the second sub-scenario is one in which we'd have a graph that looks like the following DAG.

What I'm confused about is if we condition on a level of tar in a big population, we'll still see correlation between smoking and cancer via the trait assuming there's independent noise feeding into each of these nodes. More concretely, presumably people will smoke different amounts based on some other unobserved factors outside this trait. So at at least certain levels of tar in lungs, we'll have people who do/don't have the trait, meaning there'll be a correlation between smoking and cancer even in different tar level sub-populations. That said, in the purely deterministic simplified scenario, I see your point.

Alternatively, I'm pretty sure applying the front-door criterion (explanation) would properly identify the zero causal effect of smoking on cancer in this scenario (again assuming all the relationships aren't purely deterministic).

AllAmericanBreakfast's Shortform

If you haven't seen Half-assing it with everything you've got, I'd definitely recommend it as an alternative perspective on this issue.

Why isn't JS a popular language for deep learning?

I haven't researched this extensively but have used the Python data science toolkit for a while now and so can comment on its advantages.

To start, I think it's important to reframe the question a bit. At least in my neck of the woods, very few people just do deep learning with Python. Instead, a lot of people use Python to do Machine Learning, Data Science, Stats (although hardcore stats seems to have a historical bias towards R). This leads to two big benefits of using Python: pretty good support for vectorized operations and numerical computing (via calling into lower level languages of course and also Cython) and a toolkit for "full stack" data science and machine learning.

Regarding the numerical computing side of things, I'm not super up-to-date on the JS numerical computing ecosystem but when I last checked, JS had neither good pre-existing libraries that compared to numpy nor as good a setup for integrating with the lower level numerical computing ecosystem (but I also didn't look hard for it in fairness).

Regarding the full stack ML / DS point, in practice, modeling is a small part of the overall ML / DS workflow, especially once you go outside the realm of benchmark datasets or introduce matters of scale. The former involves handling data processing and analysis (transformation, plotting, aggregation) in addition to building models. Python (and R for what it's worth) has a suite of battle-hardened libraries and tools for both data processing -- things in the vein of airflow, luigi, etc. -- and analysis -- pandas, scipy, seaborn, matplotlib, etc. -- that, as far as I know Javascript lacks.

ETA: To be clear, Python has lots of downsides and doesn't solve any of these problems perfectly, but the question focused on relative to JS so I tried to answer in the same vein.

Thoughts on ADHD

I've never been evaluated for ADHD (or seriously considered it) but some of these -- especially 2, 3, 6, 7, 9 -- feel very familiar to me.

Load More