elsherbini

Does anyone have a recommendation for a set up for spaced repition with the following features?

- Sync across devices

- Plain text representation of cards

- Nice mobile interface (I anticipate usually using my phone, but don't want to author cards on the phone exclusively, or be locked into an app)

Is the answer anki+ankiweb+ankdroid? Is there a better alternative?

I mean, for women in southern Africa, some communities have HIV prevalence of ~40% to this day. 4 in 10. Despite the lower risk associated with individual instances of vaginal sex, it absolutely is something to be worried about in those communities. And because those communities have lower access to healthcare, it often goes undiagnosed and untreated for far too long.

Some numbers for people interested in HIV:

Many countries in southern Africa have adult prevalence rates of >10% of HIV. In other parts of Africa most countries have a prevalence of ~1% or less. On other continents most countries have less than 1%, including the US with 0.3% adult prevalence. [1] For all the people wondering why this vaccine is being tested where it is, it's because just like we did for the covid vaccine, you split people up into placebo vs vaccine groups and then wait to see how many people get the virus from each group. If rates of HIV infection are as low as in the United States, you'd need to test the vaccine on ~100-1000x more people to see if it worked.

Non sequitur, but too interesting not to share: The estimated risk of acquiring HIV from a single sexual act is way lower than I realized until recently. Receptive vaginal sex with an infected individual is estimated to lead to infection only 8/10,000 times. Receptive anal sex is ~10-20 fold higher, 138/10,000. Sharing needles with an infected person is 63/10,000 [2]. That means you could have an ejaculates worth of HIV virions in your rectum, and be uninfected 49 times out of 50. (However, HIV infects immune cells, so it's way more likely to find an immune cell to infect if there are more in the area. That means infections or inflammation from other causes increase your risk of acquiring HIV compared to having no infection or inflammation. Similarly, way more target cells in blood, so if you have epithelial damage your risk goes up.)

[1] https://en.wikipedia.org/wiki/List_of_countries_by_HIV/AIDS_...

[2] https://www.cdc.gov/hiv/risk/estimates/riskbehaviors.html

You can compare specific cities and see a lot more information. If you scroll down the comparison page it has chance of precipitation and average monthly rainfall: https://weatherspark.com/compare/y/17721~45062~476/Compariso...

Yeah, what I really meant was "pre-prediabetes." Because temporally what's called prediabetes usually occurs after being overweight too. If there is a physiological switch that, once triggered, leads to diabetes and also causes obesity, then it would also explain the odds-ratio but it would not be obesity causing diabetes.

I enjoyed this article. This part felt weird though when you go on to explain the pitfalls of correlation chains:

> Now, obesity causes diabetes. That much is indisputable.

The odds-ratios you quote still just means obesity is correlated with diabetes. They could both be caused by the same thing, for instance an inflammatory diet. Or diabetes could cause obesity (temporally this doesn't seem to make sense, obese people go on to develop type-2 diabetes. But there could be a pre-diabetic state that we don't currently understand/measure which, when entered, leads to obesity.)

Some cyanobacteria can take up nitrogen from the air (well, dissolved inorganic nitrogen), but many can only use organic nitrogen, like N03 or NH4. Those can be limiting in the ocean because lots of life wants to grab it up.

Taking up inorganic nitrogen is called nitrogen fixation and takes a bunch of machinery and energy. Also, oxygen inhibits the process, so for photosynthetic bacteria a bunch of extra steps have to be taken to compartmentalize photosynthesis and nitrogen fixation. So not all cyanobacteria can do it, and even the ones who are able to will prefer to take up organic nitrogen if available, and they grow slower when they have to use N2.

Here are a bunch of different media people use to grow cyanobacteria:

https://www-cyanosite.bio.purdue.edu/media/table/media.html

Mostly they need nitrogen, phosphate, and sulfur to make protein and nucleic acids, and trace metals,ions, and some vitamins to use as cofactors for enzymes. In the wild, the limiting nutrient for cyanobacteria is often iron or nitrogen.

They don't need an added carbon source in their media since they get to eat dissolved CO2 from the air. Every carbon atom in newly synthesized molecules comes from CO2, and there are (very roughly) 10^10 carbon atoms per bacterial cell. http://book.bionumbers.org/what-is-the-elemental-composition... . So if you know the growth rate you can estimate the carbon fixation rate.

The Korg Minilogue is a really good beginner synth. It has a what you see is what you get interface, and even has an an oscilliscope which relly helps starting out. Here's a video plugging this as a great beginner's synth: https://www.youtube.com/watch?v=EE7EvKYTs4I

if you look at the recently sold on reverb, you see you can get one for ~$300 or ~$350. https://reverb.com/marketplace?query=Korg%20Minilogue&show_o...

One generalizable takeaway from this - smaller studies can only show really huge effect sizes. This is what statistical power means - to see small differences between groups you need really large samples. This is fine if you have actually have a huge effect - you probably don't need a large sample size to detect if sugar causes insulin levels to go up in people that are fasting.

But if you don't really have large effect sizes in whatever you are studying, then you are screwed. All your statistically significant results are going to have really large effect sizes, because those are the only ones you can see in your small sample size. If you didn't get a significant result, you don't publish, or you stop following up, and if you do get a significant result you publish and do more follow up studies. In time with larger studies or new technology that changes how we can measure things, it becomes apparent that you were chasing noise.

Off topic, but you seem like someone who might like these less popular music theory channels I really enjoy (all to do with harmony mostly)

Metaharmony - https://www.youtube.com/watch?v=JNKBF8Phj3M

The Labyrinth of Limitations - https://www.youtube.com/watch?v=86XSsHlJSrc

Synthase / Midnight Oil Collective - https://www.youtube.com/watch?v=-gWMW1Hkdik

My favorite way to find out what flags do is explainshell:

e.g. https://explainshell.com/explain?cmd=wget+-c

I also recommend this podcast, just finished it this week.

Another quote from the press-release of the study:

>“Epidemiological tracking shows a steep trajectory of new infections, indicating Omicron’s rapid spread, but so far with a flatter trajectory of hospital admissions, possibly indicating lower severity,” explains Dr Noach. “This lesser severity could, however, be confounded by the high seroprevalence levels of SARS CoV-2 antibodies in the general South African population, especially following an extensive Delta wave of infections.” [1].

So, to actually make the comparison we'd need to compare new Delta infections with new Omicron infections time-matched, comparing people who were seronegative to start with. This would be really hard to do in South Africa since there is a high proportion of seropositive people.

[1] https://www.discovery.co.za/corporate/news-room

South Africa has <25% vaccintion rate. "South Africa has a better vaccination rate than most countries on the continent: Just under one-quarter of the population has been fully vaccinated" [1]

[1] https://www.nytimes.com/2021/11/28/world/africa/south-africa...

I cant find a source, but I remember a discussion here on a brand of smart tvs that still phone home by connecting to other nearby smart tvs that are connected to a network, even if you don't connect yours to the network.

I'm relatively new to electric guitar and have a Tagima telecaster clone (T55pwh). I love it - it has no business being as good as it is for as cheap as it is.

One thing about telecasters that you mentioned that is probably not the best design is the 3 barrel string saddles that make intonation challenging (since you can't adjust the intonation of one string without changing another one too). My favorite telecaster player Tim Lerch has a video explaining the way he sets up his teles, and one of the things he does is change to compensated saddles.[1]

[1] https://www.youtube.com/watch?v=S5l54uQzKjk

This looks cool, I especially like the "intuitive" longest line algorithm.

~6 years ago I was trying to build a dashboard that displayed real-time measurements from a beehive. The sensors would take temperature, weight, humidity, etc. Back then I used simplify.js [0] which uses two simplification algorithms in two passes. The more compute intensive one is the Ramer Douglas Peucker algorithm [1]. One issue I had with streaming data is that new data points could change the past line, at least with my naive implementation. I'd love to see a real-time / streaming time series simplification algorithm where the past points don't appear to change wildly despite continuing to downsample.

[0] https://github.com/mourner/simplify-js

[1] https://en.wikipedia.org/wiki/Ramer%E2%80%93Douglas%E2%80%93...

Very cool that you can register font names that have styles attached. And all of the ligature support seems great for adding math expressions to your plots.

What are the pros and cons of using pymc3 vs writing a model in Stan and using pystan/rstan? It looks like the minibatch support of variational inference is one feature that pymc3 has up on Stan.

Another question, what are the practical limits of MCMC with lot? I asked a question on statsexchange a year ago [1] that had ~2000 data points, but it is analogous to a research problem I have that would have more like ~100 million data points. Would a model like the one I suggested there ever finish running?

[1] https://stats.stackexchange.com/questions/441250/pope-effect...

Those gradient names are annoying, but they are a de facto standard from ColorBrewer[1]. I always have to look at that website to figure out the abbreviation I need in ggplot2.

[1] https://colorbrewer2.org/#type=sequential&scheme=GnBu&n=3

Seconded. He has a full course on youtube as well, and a free version of the textbook that is just missing the last chapter available on his website (password is in the first or second lecture on youtube)

https://www.youtube.com/watch?v=4WVelCswXo4&list=PLDcUM9US4X...

Stan uses MCMC (specifically NUTS, which is a Hamiltonian Monte Carlo sampler) to optimize parameter fitting, so it can be used for things like ODEs.

Here is an example from a class taught last January that uses stan to fit a simple ODE (using the `integrate_ode_rk45` function in stan):

https://github.com/gregbritten/BayesianEcosystems_IAP/blob/m...

Does anyone know what a practical upper limit is for Stan in terms of size of data set / number of parameters to fit? Can you use stan to fit a model with ~10^4 parameters and ~10^6 rows of data if you had access to ~10^3 cores? How long would it take?

This was useful. Do you know how painful it would be to use Stan with 100k rows, or even 1m? (For a sorta normal hierarchical model)

Did you know it's illegal in the state of California to administer an IQ test to a black child? [1] This is a fantastic episode from radiolab about the history of IQ tests in the US and goes into how a test can be racist even without seeming so.

[1] https://www.wnycstudios.org/podcasts/radiolab/articles/g-mis...

Oh great, I need to update then! Is the beta available to test somewhere?

I have an OP-1 and an OP-Z. More and more I use the OP-Z, the sequencing workflow is just incredible. I don't have an ios device and after climbing the learning curve it's totally functional without a screen.

Also, the OP-Z can be used to sequence several pieces of external hardware at once [0]. Definitely not a toy. It has it's quirks (e.g. adding your own samples requires chopping them to exactly 6 seconds or all hell breaks loose, and no ability to "zoom in" on your sequence if you have multiple trigs per step) but it's a trade-off for all the power in the form factor.

[0] https://www.youtube.com/watch?v=XHfCl6-xamM

I like it how it is. It reminds me of Snail Drawings by Daniel Renali [0], it makes me think about order/logic and the contrast to a lot of patterns seen in nature.

[0] https://www.danielranalli.com/recent-work/snail-drawings-ser...

Glyphosate isn't sprayed on food?

edit: Of course it is sprayed on food [0][1].

[0] https://www.youtube.com/watch?v=lW8SrXMW7Ug [1] http://web.mit.edu/demoscience/Monsanto/about.html