Introduction

This is an effort to study the behavioral characteristics of Chatous. In this study we have 320,853 users.

Gender Distribution

We can see that there are far more males than females so we will sample out the males to make the distribution 50-50.

users_sam <- jusers %>% group_by(gender) %>% sample_n(sum(jusers$gender=="Female"))
table(users_sam$gender)  
## 
## Female   Male   None 
## 121461 121461      0

That is better.

Number of Chats

Here we are focusing in on the basic statistics of numbers of chats.

Here we can see there is a clear delineation that males are chatting more than females

Number of Chats Started

Now we will focus on the number of chats which the user initiated

We can see clearly that males are starting more chats than females

Number of chats terminated by the user

Since males are starting more chats than females this also seems to have the same trend.

But what if we look at the percentage instead?

Now this is striking. Females are terminating a much higher percentage of their conversations.

Lets drill down even further and have a look at the percentage of chats that a user terminates which they did not start.

Now let drill in one more time and check the percentage of times users were terminating chats without responding

This is actually interesting. It would appear that females tend to respond to a higher percentage of chats than males.

Chat Length

Note: One has been added to the length of chat as we had many chats that laster under 1 second (i.e. instant disconects)

We can see that in general females tend to talk longer than males. However, the small spike at 1 (which is really 0) of the females is alo interesting. This indicates that females have a higher propensity for instantly closing down a chat.

Here we can see that there is a huge concentration of chats by males which have 0 words in them. Also females tend to have longer conversations (word-wise) which is consistent with the pervious graph.

Here we see that males dominate the extremes of 0 and 100%. Although lumps the people who have no words at all in their conversations in the 0 bucket. Eliminated everyone who has no words in any conversations we get the following:

Once again this is striking (although not unexpected). We have many females who speak very little, while we have many males which do all the talking.

Now we will look at how fast people are talking.

I’m not sure we really get much out of this as they seem to follow the same general pattern.