Stats Made Easy

Practical Tools for Effective Experimentation

Sunday, May 13, 2007

Experimentation uncovers most desirable time to embark on morning commute

In today's Ask Marilyn® column in Parade magazine, a reader questions whether a commuter might get to work earlier by leaving home later. Beam me to my cubicle Scotty!

I’ve never considered this idea as a possibility, but I have experimented on varying times of departing for my daily commute of 20 miles to the Stat-Ease office in Minneapolis. My goal is to precisely predict the rush periods and avoid them while working within our flex-time rules for full-time workers. The graph shown here illustrates my theory on traffic around the Twin Cities (I drive through Saint Paul from a suburb to the east – Stillwater, Minnesota).* My belief, based on decades of daily commuting and not refuted by these experimental results, is that cars congregate in waves due to differences in working hours – some drivers working the 7 AM factory shift and others expected to be at their desks by 8 AM, for example. For me the most desirable departure is at 6:34 AM,** which maximizes my sleep time and hits a trough in the waves of traffic – still 35 minutes on the road for my average commute. I call it the “hole” and when I hit it right, my car is like the container you put into the drive-through receptacles of the bank – it whooshes me into Minneapolis. My statistical colleagues question my theory due to the sparsity of recorded data, so if any of you can provide support, I’d appreciate it. It boils down to this: Obviously one can leave later and get to work faster, but the trick is not to be late.

*For all the details, see the One-Factor RSM [Response Surface Methods] Tutorial for Design-Expert® version 7 software.

** Refer to the One-Factor RSM Tutorial (Part 2 – Advanced topics) for details on how Design-Expert’s numerical optimizer found the most desirable combination for leaving at the latest, minimizing drive-time and making the results least susceptible to variations in departure via propagation of error – POE.


Post a Comment

<< Home