Tag Archives: Statistics

Statistics for Human-Computer Interaction

Can statistics be used to improve human-computer interaction?

Of course!

I think it's common knowledge that the QWERTY layout that most of us use is the way it is because it was an optimal layout for typewriters. Keys were ordered in such a way as to reduce the hammers of common characters jamming.

Only today, we don't have hammers. Nor ink spools. (Nor the smell of oil after you've been typing for half an hour!)

So why do we still have QWERTY? Is there a statistically better key layout for humans now that we don't have to worry about typewriter hammers jamming?

YES.

I've started trying to relearn how to type using a layout called Colemak-DH.

Thinking in terms of touch-typing, where your fingers rest on the home row; in QWERTY, 32% of typing is done on the home row. In Colemak, 74% of typing is done on the home row. (ref: Wikipedia)

I like those numbers!

The Colemak-DH layout arguably improves upon Colemak. It reduces horizontal finger movement by changing the center column of keys further: Colemak center column usage being 14.8% while Colemak-DH is 7.8%. Big reduction. (ref: Mod-DH)

One last thing that always confused me. If you're typing on a QWERTY keyboard, have a look at the "Q", "A" and "Z" on the left hand side. They're positioned diagonally... though not perfectly. The "A" is about a third of a key along to the right, then the "Z" is two thirds away from the "A" (The "Z" isn't vertically aligned with the "Q" at all!). I can't help but think my typing inaccuracy on QWERTY has frequently been caused by this inconsistent horizontal shift. But turns out there's a way I can test this! These days you can buy ortholinear keyboards. Keyboards that have their keys arranged directly above and below each other. Some good examples here so you can see what I mean.

So I decided to take the dive. Let's look at two keyboards that give me the flexibility to type how I want!

The ZSA Voyager

The Naya Create

So in future I'll be typing a little bit about what it is to relearn how to type.

Year Two

After a long break from my first module, I decided to start my next module in February 2014, half-way though the academic year.

My second module was to be Statistics. I was looking forward to this, and wasn't disappointed. The module covered a lot about survey results, and how to conduct surveys effectively and make sense of the data. It also had a biology experiment thrown in there for good measure, the results of which were meant to be used in an assignment.

One of the things that really bothered me though, was their description of statistical variance. To my mind, one measure of variance would be the sum of the difference between the population mean and the sample mean, all divided by the sample count.

In fact, that's not the case. It's not divided by the sample count, it's divided by the sample count minus one. This seemed totally counter-intuitive, and apparently the full explanation was outside of the course material!!! Frustrated I scoured the web looking for an explanation... Eventually I came across a decent youtube vid. Thanks youtube!

To summarise, this video proves that sample variation s^2 is an unbiased estimator of the population variation \sigma^2 as shown below:

    \[<span class="ql-right-eqno">   </span><span class="ql-left-eqno">   </span><img src="https://adrianbell.me/wp-content/ql-cache/quicklatex.com-8309d8e21d5c74b4a21588f648103d5a_l3.png" height="45" width="261" class="ql-img-displayed-equation quicklatex-auto-format" alt="\begin{align*} E(s^2) =& E \left( \frac{ \sum_{i=1}^n( x_i - \bar{x} )^2 }{n-1}\right) = \sigma^2 \end{align*}" title="Rendered by QuickLaTeX.com"/>\]

See that pesky "n-1" divisor? Check out the full explanation here.

So after my study starting in February and ending in September, I felt I'd had a decent introduction to statistics.

Although, something that does surprise me is that looking ahead at my possible future module choices... statistics doesn't really crop up again for the rest of my maths degree. With the growing importance of statistics in modern society (let alone mathematics itself!) I would have expected a lot more of those modules on offer. Having said that, there is a separate BSc (hons) Mathematics and Statistics that the OU offers, but this appears to almost be purely statistics and doesn't give much variety. -certainly wouldn't be good for me.

In fact, there are  only two real interesting modules on this Maths and Stats degree course, one of which is the final year module "Mathematical Statistics" (M347) that introduces Markov Chain Monte Carlo, which is an area of interest. I suppose after my ten-year degree, if I'm still desperate for more, I can sign up for it as a stand-alone module. 🙂