Scenario: There was an online competition to win a prize. The competitors are technology start-up companies and the prize is worth around 5,000 USD.

There was a public, online vote to decide the overall winner. Competitors turned to social media to get the word out about their competition entry.

Prior to the public vote an expert panel decided which of the companies involved would make the short list. Those that made it to the short list are included in the public vote.

An interactive plot of 48 hours of the voting data collected between companies A, B, and C is available here: http://doge.bike/plot/. The source data is available here: http://doge.bike/plot/question.csv. At the beginning of this 48 hour period voting had been open for 8 days and everyone started at 0 votes.

Can you see any unusual/unexpected trends in the data? We were surprised to see the sharp spikes in the B and C lines. We thought it was odd that such large spikes would happen and then immediately stop, rather than tailing off slowly.

Static image of plot

  • $\begingroup$ Is this for a course? $\endgroup$ Commented Oct 16, 2014 at 15:22
  • $\begingroup$ @gung It's not for a course, this is a real public vote and we're trying to analyse the trends that we saw in the data. $\endgroup$
    – Matt
    Commented Oct 16, 2014 at 15:23
  • 1
    $\begingroup$ @whuber Thanks for your input, sorry I wasn't clear to start with. The y-axis is time beginning at midnight and lasting for 48 hours. The x-axis is cumulative number of votes. $\endgroup$
    – Matt
    Commented Oct 16, 2014 at 15:24
  • $\begingroup$ @whuber Is it clearer now what kind of answer/information I'm looking for? Would very much appreciate the community being able to have a go at this. $\endgroup$
    – Matt
    Commented Oct 16, 2014 at 15:29
  • $\begingroup$ Many thanks for the quick explanations. It's unclear how this is a statistical question, though: it seems to ask us to speculate about the observed behavior. Sure, like you we can look at the data and say yes, you do indeed have sharp increases followed by plateaus. But that does not represent any advance over what you have already found out. Any explanation would have to come from theories about crowd voting. $\endgroup$
    – whuber
    Commented Oct 16, 2014 at 15:30

1 Answer 1


Looking at votes per time interval may be helpful in highlighting the spikes, though it will still be difficult to model because some spikiness will be a natural result of particular social events like tweets from interested parties.

enter image description here

Distribution of the votes per 10 minute interval (A, B, C) emphasizes how A is different from the others: enter image description here

Zooming in on the big spike for C with the minute-by-minute votes, it's interesting that every minute has between 1 and 3 votes.

enter image description here

Of course, it's really interesting that the final vote was so close. Were voters aware of the running totals?

  • $\begingroup$ Thank you for your already very insightful answer, very much appreciated. Voters were aware of the running totals, yes. Would be very interested to know how that might affect your analysis/conclusions. $\endgroup$
    – Matt
    Commented Oct 17, 2014 at 14:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.