What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 134 Stack Exchange communities.

I noticed that some users on SO only ask questions while hardly answering anything, whereas others do most of the answering. I wondered what the balance between these user types looked like. For example, would the most active users be doing most of the answering? Would the less active users be those who come in only when they have a question? Such information might give some insight into the "healthiness" of the community.

With the April 2010 data dump and some Python code (below), I came up with the following plot:

Number of answers versus number of questions, per user

Each point represents a user; the location of the point corresponds to the number of questions and answers written by that user. Where multiple points coincided, they have been merged into a bigger point. Note that one user, presumably "unknown (google)", fell way outside the plot range, so I removed him.

I'm not quite happy with the result; the graph is not as informative as I'd hoped. Any ideas for improvement?


Python code: count.py and plot.py. Usage:

cat posts.xml | python count.py | python plot.py

Output will be written to a file qacount.py.

share|improve this question
1  
I gotta say I don't like the composition of that plot. Maybe I'm spoiled having used tools with good intensity plotting for 2D histograms for decades, but Blech! –  dmckee Apr 5 '10 at 23:11
1  
Kinda looks like the top view of a very large and evil food processor... –  Aarobot Apr 5 '10 at 23:44

2 Answers 2

The log-log plot tends to magnify the small numbers, where most of the noise is. What does the plot look like if it's linear on both axes?

share|improve this answer
1  
yeah, I really dislike the really obvious lines at values < 10, which took me at least 10 seconds to realize was an artifact of the log-log axis choice. It has no meaning, so it really shouldn't be emphasized. –  Jimmy Apr 6 '10 at 0:45
    
Linear makes most of the values bunch up in the lower right corner. It's not very informative either. –  Thomas Apr 6 '10 at 9:43
    
Where, with "right", I meant "left", of course :P –  Thomas Apr 6 '10 at 10:27
    
@Thomas: Try linear in answers and questions, but log(count). –  dmckee Apr 6 '10 at 14:06
    
@dmckee: Tried that too, and variations. It doesn't really help. I think I'd need to do some more bucketing, or something. –  Thomas Apr 6 '10 at 19:55

I agree with Greg that the log scaling may be hiding some salient features, but that appears to show a broad spectrum of usage patters that includes many users who engage in both asking and answering.

share|improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .