Hacker Newsnew | comments | show | ask | jobs | submit | login

We've been working hard on a more intuitive regression interface:

. Plain-English description of results

. Inline data transformations (e.g., taking a log)

. Automatic diagnostic charts and distribution visualizations

. Automatic alerts to common issues (e.g., funky residual plot)

. Plain-english guides to regression [1] and interpreting residuals [2]

. M-estimation instead of OLS

This is of course a work in progress, so we'd love to get feedback on this. Particularly if you see something in the docs you don't agree with.

[1] http://docs.statwing.com/user-friendly-guide-to-regression/

[2] http://docs.statwing.com/interpreting-residual-plots-to-impr...

(Logistic regression is also in the works, we'll probably do a "Show HN" for that later, so please hold comments on it)

-----


I've tried quite a few things and the most useful thing for me has been https://www.headspace.com/ . It's "daily" guided meditation (that after not too long is barely guided). The narrator's voice is really good, not too hippie, very calming. Everything is pleasant and well-designed.

This was also pretty good, if you'd rather read a book: http://www.amazon.com/Search-Inside-Yourself-Unexpected-Achi...

-----


Edit: Do you pay for headspace or just use their 10 minute? I just signed up.

Thanks for the book suggestion, I just bought this book.

-----


Just to clarify: I think you're saying that the best societal answer is what you suggest. I agree (as I agree with eitally).

The best if-you're-pregnant-right-now solution, though, is different, since currently very few hospitals allow birth in their physical location without a hospital physician.

The calculus in that case is probably a tradeoff of unlikely but very bad outcomes (death by homebirth) vs quite likely somewhat-bad outcomes (unnecessary Caesarean, unnecessary rushing, etc., by hospital), I'd think.

-----


Our son was delivered at a hospital by a midwife that was part of a team of 2 midwives (CPNP/CNM) and 2 doctors (OB/GYN). If things got beyond her capability she could call in the doctors for backup. Our daughter was delivered by one of the doctors due to him being on call.

So, nurse midwifes do deliver in hospitals. I actually think they are not allowed to do home births or they can lose their license.

(In the US)

-----


Wow, I'm surprised how powerful that effect is. I only played for a couple minutes (very cool), and I think it's been about a minute of my vision waving.

-----


Really good stuff, really nicely done, I love it. And good pre-Christmas launch timing :)

Small issue: I grabbed an .m4a, uploaded it, and the screen just sat there. Console errors and whatnot. Didn't take a rocket scientist to guess that .mp3 might work better (it did). So I'm all good, but just reporting it.

Again, really good stuff, congrats.

-----


Thanks for pointing this out. Right now, we're supporting wav and mp3, but I'll be adding a couple things in the very near future - greater support for various file formats, and better error handling for the ones we don't support.

-----


Agreed. From the article Nostromo just posted[1], looks like the mouse models were based on aspartame, saccharin, and sucralose, and the human subjects were taking saccharin.

[1] http://www.newscientist.com/article/mg22329872.600-artificia...

-----


This really hits the spot for me. Super pleasant. I could sit here for an hour.

-----


Thank you! That's how I feel too. It's especially fun watching it at different times of day - seeing videos from sunrise and morning commutes on the other side of the world!

-----


Same. I found myself wanting to be more immersed. Maybe the music controls could be hidden or the videos could be viewed full screen.

-----


Fullscreen is an improvement I would like to add soon. It wasn't working well during development because I actually run it with two video players - with the hidden one loading the next video so that the cutover is quick.

The videos are shown at their highest-possible resolution in the window now, though, so at least we aren't missing out on any quality (though it would be great if Instagram allowed users to upload higher resolution source videos).

-----


Also available on the iTunes store

-----


Some notes about the data, and in particular differences between how it's presented here and its raw form via Yelp:

1. Businesses can be in multiple neighborhoods in the original dataset. In this version businesses can only be in one (the more common of the neighborhoods the business was listed in). There's some nice presentation and analysis advantages to this.

2. We dropped categories with less than 50 businesses in them because of some limitations of Statwing (it slowed us down a lot without much benefit, for reasons I'm happy to explain but are pretty boring.

3. Instead of taking the number of stars typically presented on a business (1.0, 1.5, 2.0, etc.), we grabbed an average from Yelp's dataset of reviews for each of these businesses, so you end up having businesses with ratings like 1.37 or 3.22. There's spikes at 1, 1.5, 2, etc. because of businesses with very few reviews, so filtering to only include businesses with >25 reviews is pretty handy.

4. This is only one of several datasets Yelp provides (one for each business, one for each review, one for each user, etc.) http://www.yelp.com/dataset_challenge

Final note is that we're of course always interested in feedback, so have at it.

-----


> 3. Instead of taking the number of stars typically presented on a business (1.0, 1.5, 2.0, etc.), we grabbed an average from Yelp's dataset of reviews for each of these businesses, so you end up having businesses with ratings like 1.37 or 3.22.

I don't believe that derivation is equivalent.

From my own tabulation of the data:

# of reviews in Yelp's reviews dataset: 1,125,458 reviews

# sum reviews among all reviews for businesses in Yelp's business dataset: 1,236,445 reviews

So the aggregate will fail to account for about 10% of the rating data.

-----


There's definitely some inconsistency here.

An even larger issue is probably that the way Yelp calculates ratings for a business isn't a straight average, it involves a notion of a prior expectation. I'd go into more detail here but I'm struggling to find the (I think official?) URL talking about this.

-----


We have authorization from Yelp representatives to show their data in this fashion.

-----

More

Applications are open for YC Summer 2015

Guidelines | FAQ | Support | API | Lists | Bookmarklet | DMCA | Y Combinator | Apply | Contact

Search: