2
$\begingroup$

I have an archive of 20 years of email data that I am analyzing. Currently, I have grouped each email by day and hour, because I want to be able visualize how much email is sent at given hours of the day. Specifically, I am trying to answer the question, "Are some hours of the day more likely to have email than others?" My first approach was to create a line plot of the mean. Line Plot of Mean Posts per Hour

Although this gives a good characterization of the "average" behavior (and reflects my expected trends), it does not give any characterization of the variation within an hour. Just from looking at the data, I know that most days have no emails sent at all, so most of the time the data is zero (in fact, plotting the median results in a horizontal line at 0 posts). The distribution for each hour is definitely not normal; there is extreme skewing towards zero for every hour.

My next thought was a scatter plot with high transparency, plotting each day's # of emails for each hour. Although better, I still think this struggles to convey the story of the data meaningfully:

Scatter Plot of Posts Made over Hours

Does anyone have better ideas on how to accurately capture the data in a visualization? To me, it seems like there are clear variations during the day, but most days have zero posts.

$\endgroup$
2
$\begingroup$

I've ended up using a 2D Histogram for this (basically a heatmap). It seemed to work pretty well, although we'll see when I start showing it to my colleagues :)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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