delib

Some of my gripes with Yazio:

- It freezes. all. the. time. For an app that I use 10+ times per day that's just not acceptable.

- When I update a shared recipe on my phone, it doesn't update for my wife. The only way to solve this is log out of the app and log back in. When you do this, you realize that not all your settings are saved, so you need to go back and fix those.

- They seem to use some aggressive caching in their UI: For instance, I update the amount of food, swipe back, and still see the old value.

- Can't do anything without an internet connection, not even look at how many calories you have left.

- Newest bug: Shared recipes don't open anymore. Just switches to the recipe screen. When I reported this to them, I got an AI answer on how to create recipes.

There are other funny things: For a few weeks now, you can collect "diamonds" when you track foods. I now have hundreds of these, but can't do anything with them.

I want to try that eventually, but I feel like I'm not quite there yet with just going on vibes.

The CI is for the mean, not for the entire data. This is the difference between a confidence and a prediction interval: https://stats.stackexchange.com/questions/225652/prediction-...

I do agree that it's fairly noisy though.

The concept is a bit confusingly named, but that doesn't make it "bogus".

(Period) life expectancy at birth summarizes the current health status of a population, nothing more, nothing less. It doesn't make a predictions about what happens in 2078.

Yes, the target does not need to follow a normal distribution, but the target conditional on the features (which is what the original authors wrote) follows a normal distribution if you make the assumption that the errors are normally distributed. The two statements are equivalent for the standard linear regression model.

it's linear regression on the log-odds

Almost - logistic regression assumes that the function is linear in the log odds, i.e. log(p/(1-p)) = Xb + e. The problem is that you can't compute the log-odds, because you don't know p.

Not all are: https://cdn.shopify.com/s/files/1/0644/8811/products/Staedtl...

This is not the only reason, though. The U.S. spends a higher percentage of its GDP on defense than other countries.

I think an overlooked aspect of the German-English comparison is the system of tenses. In German, you can get by with basically just two tenses: the present and the perfect. The past tense can almost always be replaced by the perfect when speaking. When you want to express something that lies in the future, you can just use the present and it will in most cases be clear from context.

Compare that to English, where the mastery of the tenses (i.e. when to use which tense) is essential. Distinguishing the use of the simple forms from the progressive forms can be tough for non-native speakers. And then there are all the different past tenses: simple past, present perfect, past progressive... in many cases it sounds really strange when these get mixed up. Think of something like: I have gone to the store last night. Or: He never was in New York. (When the speaker means that he has never been.)

I scratch my head at a society that makes its members memorize such useless information.

Such a weird perspective on language and on the agency of societies. There are tons of "useless" things in any language. For instance, why not get rid of all the tenses in English? Do we really need the past progressive?

To make English spelling more "logical" (so people need to remember less useless information!), we should again look to Mark Twain:

Fainali, xen, aafte sam 20 iers ov orxogrefkl riform, wi wud hev a lojikl, kohirnt speling in ius xrewawt xe Ingliy-spiking werld. (http://grammar.ccc.commnet.edu/grammar/twain.htm)

For less structured diagrams, I really like Scapple: https://www.literatureandlatte.com/scapple/overview

I use LyX to write academic papers and really love it. Sure, there are quirks, but it's still makes for a better workflow compared to the alternatives.

Especially important for me:

- track changes, which makes collaboration convenient (especially with less technically-oriented colleagues)

- math typesetting with instant preview

- integrated BibTex support, which makes citing easy

- easy to reference equations, tables, and figures in the text

people that simply feel something about a tiny corner of the whole picture

Why is this about the author's feelings? There are plenty of links in the piece to underlying data sources, and these show that the problems mentioned don't vary a lot by discipline. For instance: http://phdcompletion.org

be much better supported by typical social science approaches

This implies that linear regression is not a typical social science method. If you review articles in sociology or political science journals, you will find that many articles use quantitative methods, and nearly all of those quantitative articles use regression.

They also attempt to get at the causality implied in their hypothesis. It's hard to do that purely with qualitative methods. Not everybody might find their IV regression convincing (they rarely are), but it's good to see that they tested their hypothesis with data.

Here's a map showing where it's quicker to walk: http://rodcorp.typepad.com/rodcorp/2003/10/london_tube_map.h...

Right (although this depends on the journal). But these problems are a function of Elsevier's immense market power as well.

Elsevier does provide a service.* They manage paper submissions, format papers, sometimes edit them, publish them, etc. Without Elsevier (or other traditional publishers), this work has to be done elsewhere. It thus may make sense for universities to outsource the publishing to a dedicated company. However, Elsevier has become big enough to dictate the terms of the contract. The German universities are doing a very smart thing in pooling their market power. They don't want anything for free, but they want to pay a reasonable price.

* Edit: Of course, they don't do it in a particularly good way.