Tweets per day

The amount we tweet per day can change over time. This data can be really noisy: On some days we might hammer out one tweet after the other, while on other days we don't get around to doing much tweeting at all. There might be vacations, days where you're sick in bed etc. This can make it very hard to see long-term trends.

For this reason the data here was averaged to remove the noise. This was done by applying a rolling average on a 180 day window. This should make it much easier to see how one's behaviour changed over larger time-frames (In pandas this rolling average is created by dataframe.rolling('180d').mean() ).

In our graph to the left the x-axis gives us the time since one signed up for Twitter and the y-axis gives the normalized number of tweets on a given day.

Different types of tweets over time

There are different types of tweets one can send on Twitter. Some tweets are replies to tweets by other users, while other tweets are retweets of tweets done by other users. Last, but not least there are the good old regular tweets, which are neither a reply to anyone nor a retweet of tweet.

Again, this data might be very noisy due to short-time changes. For this reason the data was averaged to remove the noise as well. This was done by applying the same rolling average on a 180 day window as above. This should make it much easier to see how one's behaviour changed over larger time-frames (In pandas this rolling average is created by dataframe.rolling('180d').mean() ).

In our graph to the left the x-axis gives us the time since one signed up for Twitter and the y-axis gives the percentage that different tweet types account for on the overall tweet volume.

When are tweets posted?

Different people have different styles in how they use Twitter. Some are owls and tweet deep into the night, while others of us are larks and send our first tweets out while the sun rises. And of course, there might be differences between workdays and weekends.

Twitter does not save the local time in which a tweet was sent. Instead it gives all dates and times in UTC. To get the correct local time it is thus necessary to know the latitude/longitude from which a given tweet was sent. Correspondingly our graph to the left is based purely on those tweets for which a geo location is known.

In our graph to the left the x-axis gives us the hour of the day in which tweets were sent the y-axis gives us the overall number of (geotagged) tweets that were sent on a given workday or day of the weekend.