#21

@Max, is that pdf available anywhere else, please?

I just get an error with your link, which may be due to dropbox being blocked at work (we'd normally get a warning message though).

I am interested in how noise can be represented to a less technical audience.

#22

Can else access the pdf? (EDIT browser refresh says yes)

#23

I can, but it was redirected through dropbox which may be disabled @martin.R's workplace.

#25

Here you go (757.8 KB)

(Note to those using attachments. Even though the upload progress is 100% and you can choose "reply", wait a while longer until you see the link ur resolve and it turn to html. Otherwise, you link text will be "Uploading...". Not. Cool.)

1 Like
#26

Dataisbeautiful is particularly unpredictable for what they decide is worthy of an upvote. Success and failure on really tend to hinge on how the subreddit algorithm is feeling that day.

My most noteworthy (according to the internet) project was when I used R to win an argument with my wife. Ended up getting about a half a million unique hits when I hit the front page of reddit. Totally decimated my website, so I wrote a follow-up post about how I moved hosting providers and used R to figure out if the new cheap hosting solution would be able to handle the reddit front page. Data analysis isn't nearly as interesting without the human angle, so I haven't been able to test it out yet. Here's hoping for another interesting spousal argument!

In terms of ongoing projects, I have been developing an open source R package for generating and evaluating optimal experimental designs, skpr. It's a package that would be totally impractical to easily implement in anything but R, but was possible thanks to the extensive built-in statistical support and amazing frameworks and libraries (Rcpp, RcppArmadillo, Shiny, testthat) provided by the R community. Building a quality package and getting it on the CRAN is certainly an interesting journey.

2 Likes
#27

I really enjoyed this @mara, I hadn't come across it before! What did you use to create it?

1 Like
#28

I worked with Elijah Meeks, who's some sort of a D3 magician, so there are pieces of the how I didn't fully understand. But, R for processing (we hand coded things, so it was a huge task), and then React (which Elijah had to really hold my hand through). There NLP stuff, though, is all R— mainly @julia & @drob's tidytext :package: and :closed_book:!

1 Like
#29

I had a lot of fun with Baby Names. It wasn't complex, but rather, a proof of concept for a bunch of students I was teaching.

My advice to them was to build a portfolio to show their skills, and their complaint was but where can I get data? Data is everywhere you child! Find something you're interested in, and find the story. That'll be far more compelling than how well you scored on a kaggle dataset.

As proof, I wrote up that piece in one day. Nothing more than a name, a gender, a year, and a count. Four measly columns, tons of untapped potential. Now stop whining and go show me something.

2 Likes
#30

Thanks! But no, sorry, I never posted it elsewhere.

I did think it'd be fun to post it someplace more mainstream for R users, but I'm not a blogger and the dataset was too big to use something like RPubs for.

1 Like
#31

Thanks @mara. I've considered learning D3. It seems like a powerful combo with R for interactive visualisations. (You've also made me realise I should watch Archer again).

1 Like
#32

With the caveat that I definitely would've done very little "bespoke" js without Elijah (I don't think a "radial bump chart" was a real thing until Elijah made it one :joy:), there are a ton of great D3 × R resources by @timelyportfolio to check out!

Sir Portfolio's also working on porting Elijah's new framework, Semiotic, to R.

2 Likes
#33

The saying was that if there was a js library you wanted to use in R, wait 24 hours and Ramnath would build a package. @timelyportfolio is the heir to that phrase. Anytime I see something js related I check his github.

2 Likes
#34

My current favourite project is the statistics page I made for my fantasy NBA league. Originally I only attempted to recover historic results for the league but it turns out people are much more likely to sift through their old e-mails when the results will be put on a semi attractive webpage rather than collected into csv files.

Here's the link if anyone wants to see a bunch of weird European names:
https://martj42.github.io/nba_fantasy/

1 Like