My other recent data collection project had to do with the elevators in my building. It’s only 13 floors, but it has five elevators. Seems like a lot. More than a third of the time, you’d expect one to be waiting for you on your floor, though they never are. I started noticing that two of the elevators rarely if ever catered to my vertical transportation needs. So I used Counter to collect some data to see if, indeed, those two elevators were less likely to pick me up.

I plugged the data into PSPP and ran a chi-square analysis and, indeed, my data is only 0.03 percent away from giving me 95% certainty that the last two elevators are less likely to be the ones opening for me (p = 0.0546).

However, a funny thing happened partway through the data collection period. For the first several weeks of data collection, the last two elevators never open for me, not even once. Limiting the analysis to only this portion of the dataset shows with near 100% certainty (p < 0.0001) that the last two elevators were opening at a different rate than the other three.

But then one day I pressed the button and, voilà, elevator four opened. Ever since that day, the last two elevators have been appearing at the same rate as their brethren (p = 0.9613)…

…making it perfectly reasonable to conclude that whatever was wrong with elevators four and five has been fixed.

One interesting dynamic today with statistics tools is that there are lots of specialized statistics calculators out on the web today. The advantage of a specialized online calculator is that the interface can be customized for that kind of analysis. At times, particularly with a quick analysis of a small dataset, this can be a lot easier than feeding the data into a generalized tool like PSPP.


3 thoughts on “More PSPP

  1. plilja

    Most modern elevator systems use some kind of predictive algorithm to figure out what floors to have cars wait on between calls. The change in frequency could even have been due to your calling the elevators over time. The last two elevators perhaps were staged for a typical pattern on another floor at the start of the observation period, but since you kept hitting the button every day the system decided it better keep those cars fully in service in case you rang…perhaps an unintended observer effect (although probably not…)

    1. robertmulcahy Post author

      I was thinking a little about this the other day and whether with 5 elevators for 13 floors whether the elevator shouldn’t be waiting for me more frequently. I guess 13 floors means there are 25 possible positions for each elevator (either at one of the floors or in between). With 5 elevators, that means with no optimization I would expect that 20% of the time the elevator would open right up for me after I press the button. That sounds testable…

      1. robertmulcahy Post author

        After a couple of weeks of data collection testing the hypothesis that at least 20% of the time I should just be able to walk right on to an elevator in my building, fully 50% of the time an elevator is right there ready to lift me to sky when I’m on floor one (typically in the early morning). And when I’m on floor 11 (typically in the later afternoon), the elevator is waiting for me 23% of the time. Data collection continues.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s