| About Me

90 Days of Learning

An area I’m keen to experiment with batching strategies for is learning. Specifically getting “over the hump” with things that require a substantial upfront commitment before a return is delivered. Today I'm kicking off a 90 day experiment of batching together learning three such things.

The three examples I'm focussing on are:

  1. Core statistical methods applicable to data science
  2. Kite Surfing
  3. The RYA Day Skipper qualification

Those who don’t know me may not see an immediate correlation between those three topics. Those who do may recognise them as things I’ve spent a disproportionate amount of time talking about in the last 18 months vs the amount of tangible progress I’ve actually made on them.

My working theory is that each of them require a substantial amount of focussed investment before it becomes useful and while the definition of focused attention varies between them, in all cases it’s of a type that lends itself to batching rather than reactive habits.

The Principles

The experiment is to spend initially 30 days where my primary focus outside of work is to get "over the hump" in these three disciplines, with an expectation that if it looks to be working, I'll extend it for an additional 60 days.

The principles are less "rules" to be enforced, and more so that if the experiment is succesful, I can replicate in future, and if it's not, I have something to look back and ask "what didn't work".

  1. In any given week, some progress towards must be made towards one of these things at least 5 days out of 7
  2. Progress must be documented, either through TIL or blog posts
  3. Progress is of course subjective, but generally should involve at least an hour of effort
  4. Other regular goals will be relaxed during this period, specifically target for gym attendance will go from 5 sessions per week to 3

The core statistical methods in data science

Whilst I have a comfortable grounding in statistics from both education and previous projects, I’m keen to increase my level of practical expertise.

To give a specific example; if a data scientist explained to me that a model was written in R, utilised principal component analysis followed by multiple linear regression, I would understand what that means and be able to have a sensible conversation about limitations and other approaches.

If, however, I had a data set with multiple dimensions and thought “I wonder what the principle components of this are”. I wouldn’t be able to jump into R and calculate this myself. Similarly if one of our data scientists shows me a model in R, I can generally understand with their guidance how it works, but my ability to play with it and make small changes to it myself are limited.

So my goal is to develop sufficient competency in basic statistical methods in R, to be able to perform my own exploratory analysis and readily comprehend similar analysis from others.

Previous attempts have centred around the book Practical statistical methods for data scientists and have followed my default approach of “find half an hour to an hour every day”. While I've found the book itself to be excllent, these have generally failed not due to the learning itself but due to lack of momentum or to use a (catchy) acronym I made up, due to a high TTTIFLP.

TTTIFLP (pronounced "tifulp") or “time to this is fun let’s play” is a metric in the loosest sense of the word which looks at how long until you gain enough knowledge to enjoyably explore in a self guided fashion.

As an example Ruby in Rails has an incredibly low TTOFLP, anybody with the most basic understanding of programming can go from never having used it to building fairly comprehensive and functional web applications in a very small amount of time.

Working through practical statistics for data scientists has a fairly high TTTIFLP. While the individual concepts are fairly intuitive and quick to grasp, I've never managed to cross the tipping point where, when I've got a piece of analysis I'm interested in, I'll reach for R rather than a more familiar combination of SQL / Ruby for munging and Excel for analysis.

So my success metric for this piece of learning is to be sufficiently competent in R and the methods learned to generally prefer to conduct my analysis here rather than in the tools I already know well.

Kite Surfing

Of the three areas, kite surfing is probably the one with the greatest different in the amount I've talked about vs the amount I've actually progressed.

In Christmas 2018 I had a 1.5 hour private lesson in fairly high wind conditions in Perth WA. This was beyond fun and got me as far as body dragging; this means being in the water and attached to the kite but rather than having your feet attached to a board, you just control the kite in such a way that you are dragged through the water in a controlled fashion.

This confirmed that in one respect kite surfing has an extremely low TTTIF (notice the missing "LP"); I think I started grinning at about minute 10 of my first lesson and didn't stop for the rest of the day.

The challenge is the "Let's Play" part. Kite surfing involves a substantial amount of equipment, setup and technical skill. While a realtively safe sport when done properly, it can also be relatively dangerous, to onself and those around them, if not done properly.

So there's a certain base level of ability needed before it's practical to go and rent equipment and "play".

In Spring of 2019 (turns our to be a highly affordable if incredibly cold time of year to Kite Surf in the UK, thanks @TravisLeeStreet for giving this a go with me) I tried a one day group course at Camber sands, which while also great fun and well taught, the group nature means I didn't actually get much closer to being able to practice independantly.

So the goal for this experiment is to get to a point where I'm competant to rent equipment and practice independently. My initial strategy for this is to focus entirely on a smaller number of private lessons. Having chatted to one of the well respected kite surfing schools in the UK, they seemc confident that the fastest approach to this is 3 x 3 hours of 1to1 tuition with radio helmets to minimise the amount of time spent stopping to review progress.

The RYA Day Skipper Qualification

The RYA Day skipper is a qualification which confirms the holder is competant to skipper a yacht on coastal passages in daylight. While it is a UK focussed qualification, it is recognised internationally holding this or an equivilent qualification generally the basic requirement to be able to charter a sailing yacht anywhere in the world.

The course has two core components, a theory and a practical.

The theory test can be taken online and general estimates seem to be that 2 weekends and some evenings are sufficient, with the official RYA estimate being up to 40 hours of study may be required (so around 5 full days). The theory course needs to be completed before the practical course.

To date I've struggled to stick with studying for the theory course, primarily because having spent a fair amount of time crewing on both dingies and larger boarts growing up, the study is a mixture of things I already have some experience with, and bits I've missed completely as a result of having learned primarily through osmosis.

Combined with the fact that while interesting, there's currently no practical places to put the knowledge learned to good use, I've tended to lose focus and revert to learning things with a more immediate return.

My goal for the full 90 day experiment is to complete both the theory exam and - if feasable - the 5 day practical exam. If this is not faesable, e.g. due to course availability or weather issues, then the secondary goal would be to at least have the practical booked.

Most updates on progress will be on Twitter @TalkingQuickly or my Today I Learned.