In a previous post, I wrote about using probability binning to create categories out of numerical data. This simplifying approach works well on various kinds of distributions.

In this post, I wanted to show an example of how well it can work.

I spent about a year learning Python when I first got back into coding. A lot of people have asked me where to begin. Here’s a guide to what I found useful (or not) and why.

1. Learn Python The Hard Way by Zed Shaw.

I started here. I did the free online version.

I like the advice he gives you to just type things in, even if you don’t know what’s happening yet. This required a bit of an adjustment to my usual mindset, where I question everything all the time.

Once upon a time, I looked at a couple of data sets that involved geographical data. I wanted to actually plot the data on a map, so I had to do some shopping around for easy ways to do this quickly with python.

What I wanted to do

One of the data sets provided zipcodes, which I was able to convert to latitude and longitude (more on this below). At a minimum, I knew I wanted to be able to plot: a) location, b) a number value and text label, c) multiple colors designating groups of data points, e.g. ‘high’ or ’low’ values for a particular variable, which ideally could be toggled on and off.

More than once, and probably not for the last time, I have done a technical interview for which I was underprepared. I feel like no matter how much I try to prepare, I am always underprepared for technical interviews.

I’m going to tell you about a time I was underprepared for a few reasons, including:

a) It was the first interview where I was asked to write more than a couple lines of recursive code

Inspired by a couple of great posts by Michael Gibson, I want to talk about what robustness means to me, as someone coming from a science background.

I started working in a “wetlab” doing benchwork cancer research when I was in high school (I was 16). I learned early on that mistakes are:

• normal
• avoidable
• necessary.

That may sound paradoxical, so I’ll explain. Some failures will always happen: the phone rings and you drop something. That can happen to anybody. But you can take safeguards to prevent disaster, like making sure all your tubes are labeled and capped tightly, so if you drop them, nothing gets lost. And some failures are informative, like when (notice I don’t say ‘if’) an experiment doesn’t work out the way you expected. Mistakes will be always educational if you designed your experiment correctly.

Over the years, I’ve attended a lot of technical (read: software) talks. Once upon a time, I attended PyData, at Facebook. I’ve noticed at all of these events that the quality of presentations varies widely.

If you’re like me, and you go to technical meetups and conferences around the Bay Area, for example, you’ll notice that some talks are well-rehearsed.

But some seem like they were thrown together at the last minute.