Is it better to approximate the average number of persons in a (large) chat with a sine function or using a sum of normal distributions?

@allo Mmmh. I think for most groups there are some events where paople join or leave and otherwise there is little change. I would use an exponential function as the distribution after every event. Depending on the event it could also be a normal distribution. But I don't think have any ideo on how to model/predict when events will happen.

@Segebodo Sorry, maybe I should have added it. I am thinking about the graph for, e.g., seven days.

When plotting it, one can notice a something that looks like a sine-wave. But most people probably do not decide when they like to chat based on daylight, so there is no reason for it to be a sine.

It becomes even more complicated, when one assumes discrete events, like, e.g., a part of the people leaving exactly at 5pm instead of normally distributed around 5pm.

@allo are you 'measuring' the amount of people actively chatting or messages or sth else?


@Segebodo Currently I am only thinking about people being online.
I guess people being active might be correlated, but I am not sure.

This also depends on the type of chat. For example, a xC3 chat has a looong tail of people still being online after the event, but nobody is chatting anymore.

Sign in to participate in the conversation – a Fediverse instance for & by the Chaos community