Due to my incompetence at Web2.0, I killed my access to my Cornell Facebook account. Being somewhat remiss at losing my friends, as well as borderline codependent, I decided to re-add my friends. The distribution of confirmations was interesting enough that I decided to do this mini-lab report. As it turns out, I have more Cornell FB friends on the West Coast than I thought, which meant that people were confirming well past what I would expect for Ithaca residents. I am utterly ashamed that I did this analysis.
Background:
Since the dawn of human history, technology has outpaced the average human’s ability to manage its use. The failure is sometimes in restricting its use to humane, socially beneficial applications. But more often than not, the failure is one of individual incompetence. As an undiagnosed but suspected mediocrity, I managed to assign my Cornell email address to my Claremont Colleges Facebook profile, not realizing at the time that it would remove access to my existing Cornell profile. After a couple panicked minutes, I decided I would go ahead and attempt to re-friend individuals from that list.
Given an absence of other productive endeavors, I decided to try my best to analyze this pointless data set using my rusty scientific knowledge. I test the following hypotheses.
Hypotheses:
H0: There is nothing interesting in the distribution of times. All my friends are night-owls, and would have a uniformly random sleep schedule.
H1: There would be a drop corresponding to normal sleeping hours on the East Coast, presumably because most of the Cornell network continues to live in the Eastern time zone.
H2: There would be a drop in the period corresponding normal sleeping hours on the West Coast AND East Coast, because of greater than anticipated geographical diversity, either because of graduation or an underestimation of the number of West Coast friends who decided it was cooler to affiliate with a Cornell profile than a Harvey Mudd profile.
H3: I would drop this project before I finished it, because I would recognize that it was a complete waste of time.
Methodology:
I calculated bin size using the formulation provided by Freeman and Diaconis (via here)
A statistically robust width W bin size is calculated as
W = 2(IQR)N^(-1/3),
Where IQR is the interquartile range (75th percentile minus the 25th percentile, or 2*0.6745(s), where (s) is the standard deviation - don't know how to do greeks in html), and N is the sample size.
Time t = 0 was estimated from the timestamp of my initial post about destroying my access to my Cornell account. (Therefore, t = 0 corresponds to 10:40 pm EDT, +/- 1 min)
Results:
Key stats on the set:
N = 101
Range = [8, 1580]
Mean = 613.37
Median = 407
Standard Deviation = 506.92
W = 293.68
Figure 1 illustrates the distribution of confirmations as a function of time. Note that there isn’t a significant decline associated with sleeping hours for typically East Coast residents, though a trough appears when any sane American ought to be asleep. The subsequent rise corresponds roughly to East Coast residents struggling out of bed and heading to work.
Analysis:
Despite my best efforts, I did complete this study, so H3 is rejected. Crap.
Based on these results, H0 appears to be strongly rejeceted. Furthermore, a closer analysis of the geographic distribution of the times indicates that, indeed, I do have a fair number of friends on the West Coast. This lends support to H2:
H2: There would be a drop in the period corresponding normal sleeping hours on the West Coast AND East Coast, because of greater than anticipated geographical diversity, either because of graduation or an underestimation of the number of West Coast friends who decided it was cooler to affiliate with a Cornell profile than a Harvey Mudd profile.
I also note that the trough does not go to zero. I have a few friends still at Cornell who wake up at ungodly early hours. (Ann, I’m looking at you.)
Conclusions:
1. There was a drop in the period corresponding normal sleeping hours on the West Coast AND East Coast, because of greater than anticipated geographical diversity.
2. More research (funding and/or friends) is needed to refine the experiment to see whether we could determine a significant difference between this population and a modeled East Coast night owl population. This is important because it could just be my friends don’t go to sleep at a reasonable hour, being, you know, grad students.
3. I need to find something to do.
Acknowledgements:
This work was supported by my mom, who made me a delicious breakfast.
Contact:
You can contact this author at ryan.yamada@gmail.com
No comments:
Post a Comment