Milan's pages

Hello, I’m Milan!

As the borderline pretentious title of my website suggests, I like making pretty things: Pretty models, pretty visualisations, pretty teaching materials and websites, taking pretty photos, doing pretty climbs, going on pretty bike rides…

Actually, I kind of think that whatever we believe in or pursue ultimately boils down to what we find æsthetically pleasing. You can tell I’m not a philosopher.

More about me

Photo of my wind-swept self on the Brighton seafront

Tools I use

Hover over/tap on icon for details

Data science

Statistics

Visualisation

Machine
learning

Web
scraping

Programming

JavaScript

Python

Teaching

R Markdown

xaringan

Web design

HTML

CSS

Hugo

Audiovisual

By far not a power user but I have dabbled in these Adobe products

Lightroom

Illustrator

Premiere Pro

Audition

Things I've made

Data Analytics

Getting, processing, and analysing data is only part of the process for a data scientist. You also need to present the results to your audience in an engaging way. Eye-catching interactive analytics encourage viewers to spend time exploring data and gain insights.

Caffè neRd

As an intern in a quirky café you can only serve drink by querying the correct item from the menu using code. This gamified tutorial will walk you through the basics of data wrangling using R's tidyverse dialect.
Not optimised for mobile devices.

Escape room

Do you dare enter the room of data horrors few have managed to escape in time? Use your data-analysis skills to make it out of this ever-so-slightly thrilling escape room.
Co-authored with Jennifer Mankin during my time at Uni Sussex.

Interactive visualisations

Cute little visual explainers of statistical concepts I would use when I taught stats. This one shows how the correlation coefficient r can be understood as the cosine of the angle formed by two OLS lines of best fit, one predicting x from y and the other predicting y form x.

More viz

More to come...

My musings

Although I'm not the kind of person who needs to broadcast their thoughts, there are a few things I have Opinions™I stole this stylisation without shame from the wonderful Dr Jennifer Mankin. Go show her some love! on. You have been warned...

Making an interactive data visualisation, Part 1

Published on 28 Jul 2022

I’ll start this post with a little bit of background on why I made the visualisation in the first place. Now, I am really into climbing. Though I’m not particularly good at it, I do it quite a bit. I’m into it so much that I’ve been watching pretty much all the top-tier climbing competitions for the past few years. So, when I decided to find some data and do something fun with it in order to build up my data science portfolio after I quit being a uni teacher, climbing data were the obvious choice. I didn’t know what kind of data I would be able to get or what I would do once I have them but I was quite excited. The fun thing about projects like these is that you end up not only exercising your creative, design, and analytical skills but also learning something new about the subject matter. For example, I always knew that Akiyo Noguchi was pretty amazing, but playing with the app made me learn more about quite how amazing she was1. Finding Data There is no one right way of going about a project like this. You can, if you wish, start at the drawing board and design the end product and then figure out how to make it.

Making an interactive data visualisation, Part 2

Published on 13 Aug 2022

We left off in the previous part in a place where we ended up with quite a lot of IFSC bouldering world cup data scraped from Wikipedia and saved in a CSV file. Today, we will talk about data cleaning and wrangling, two steps that are integral to most data science projects. To spice things up a little, we won’t be using the CSV file from Part 1 but, instead, we’ll use a file including all world cup and world championships results since 2004. The data set is stored in JSON1 format and is pretty messy! I will be performing the data-wrangling in R using several packages from the tidyverse dialect. Of course, you can, more or less easily, do the same thing in any other language but I think R/tiydverse provide a pretty neat toolset for this kind of data processing. A little disclaimer This post only covers a part of the data wrangling I had to do to make the visualisation. The reason for this is that, in order to walk you through the thinking and coding behind the entire process, the blog post would have to be unbearable long. It’s already a pretty long read as is.

Introducing `geom_terrorbar()`

Published on 1 Jan 2023

Motivation I was reading Adam Mastroianni’s excellent blog post on peer review (it’s really good, go check it out) because it’s Christmas time and I don’t have to be a code monkey for a while. To give an example of the kind of rubbish that somehow makes it through peer review, Adam calls to his aid a tweet many of you will have seen but for me it was a revelation: Wait a second, these are not real error bars … the author literally just put the letter “T” above the bar graphs 😭 pic.twitter.com/KKtTGRHFaw — Josemari Feliciano (@jmtfeliciano) November 28, 2022 Going by the lively discussion in the thread, many people seem just as appreciative of this galaxy-brain-level of problem solving as I was. Now, I’d heard of Advances in Materials Science and Engineering, the journal that published the now retracted paper that this figure is from, on the Everything Herz podcast hosted by the young researcher of the year Dan Quintana and the curmudgeonly self-indulgent academia survivor James Heathers (again, go check it out!) but I wasn’t ready for quite this level of hilarity. I simply think that having Ts for errorbars is wonderful and everyone should start doing it from now on.

All posts