What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 131 Stack Exchange communities.

In discussions here on MSO, and through my own anecdotal observation, it is clear that different tags on questions

  • attract more or less interest in the question (based on views)
  • higher or lower scores for the question

So, I thought I'd graph it out. Description:

  • x-axis: the average score the question received (up votes + down votes = score) for the tag
  • y-axis: the average number of views for the question for the tag
  • z-axis (bubble size): average score / average views ("vote per view")

alt text

My two quick observations?

  • scala is the most effective vote per view tag
  • scala tag questions have the lowest average views
  • subjective is best at attracting both views and votes, and places second in the vote-per-view efficiency
  • asp.net is the least effective vote per view tag

Comments? Questions? Want another tag listed? Want the z-axis to represent something else? Want a magic pony to send you an MSO sticker?

share|improve this question
    
+1 Good work. Nice observations. –  Jonathan Sampson Sep 3 '09 at 11:30
    
average views / average score is "views per vote", not "vote per views"---which one is it? –  balpha Sep 3 '09 at 11:41
    
Yes, i seemed to have mixed the two. Let me correct that. –  Stu Thompson Sep 3 '09 at 11:46
1  
I can't really tell the difference between C++ and ASP.NET. I don't know if you have Red and Green together on this chart, but in case you do, could you change it? Some of us are Red/Green colorblind. –  George Stocker Sep 3 '09 at 13:22
1  
Can do. Shall I banish red shades? Or? More guidance please. –  Stu Thompson Sep 3 '09 at 13:44
    
I've updated the graph with different colors and other visual tweaks. Hope this does the trick for you all. I'll be sure to keep this in mind in the future. –  Stu Thompson Sep 3 '09 at 18:17
    
Nice update, Stu. Are then generated directly from your queries? Or are you doing these some other way? –  Jonathan Sampson Sep 3 '09 at 18:55
    
This graph is generated by 0) a previous import of just questions from the data dump's posts.xml; 1) a simple query (SELECT avg(score), avg(views) FROM questions where tags like '%waffels,%';); Then 3) save the file as a CSV file; 4) import it into MS Excel (on my MBP); 5) make a bubble chart from the data; and finally 6) save it as an image. Other graphs I do have been similar at times, but not always. I'm maxing out what I can do with Excel and need to move onto something more serious, like STATA, MatLab, whatever. –  Stu Thompson Sep 3 '09 at 20:31

4 Answers 4

It's interesting but one of the issues is your axes are highly related. Like average views is a component of the y and z axes.

What might be interesting is seeing these three:

  • Average views
  • Average question score
  • Average answer score

You could probably even eliminate question score (and possibly put in something else) because I think we've fairly well established that question score doesn't have much meaning.

share|improve this answer
    
Yea, the z-axis is derived from x over y, so it is really both axes that it is related too. How about number of comments or answers for the z-axis? (It is easy to do, and hence I am more likely to actually do it.) Number of tags and average questions answered percentage are easy too. –  Stu Thompson Sep 3 '09 at 13:41

subjective is best at attracting both views and votes, and places second in the vote-per-view efficiency

No. That's incorrect. You have no facts to prove this. You can only prove that they are linked. Think about it: Is the tag making the question popular, or is the question popular on its own? Basically, is the subjective tag the cause or a side-effect.

share|improve this answer
1  
I don't think it's the subjective tag at all, but the subjectivity of the questions that make them popular. It's the bike shed effect. Everyone can have an opinion on wide open subjective questions. You have a much smaller group of people looking at c++ and java questions than subjective questions. –  Bill the Lizard Sep 3 '09 at 13:37
    
I can see what you are getting at. How would you rephrase what I have above? –  Stu Thompson Sep 3 '09 at 13:42
    
I don't think anyone is going to be confused into thinking that the tag itself is generating views and votes, that's absurd. Obviously, the question itself is generating views and votes. What's demonstrated is that lots of questions that generate many views and votes happen to be tagged subjective. The assertion is that it's the subjective content that is responsible both for the stastical properties observed and (more mundanely) for the presence of the subjective tag itself. –  ベレアー アダム Sep 3 '09 at 16:07
    
My prose = repetitive much. s/itself/waffle/, that should make it more palatable. –  ベレアー アダム Sep 3 '09 at 16:09

That is completely normal.

Subjective has appeal across all technological areas. People are likely to view questions that they understand, or are related to their area of interest. Everyone understands and can comment on subjective. And because the subjective questions are usually appeals to some situation that a lot of us can relate to ("What is the best time of day to code?") a lot of people will think it's funny and upvote it.

Because Scala has such a low number of views (the lowest) it means if there are a few (2/3) Scala fiends out there who upvote everything to do with Scala, then the Scala tag's upvote to view ratio will get huge. It is easy to skew that because the population of the sample is so small. If some person on meta were to go in and downvote all the scala tagged questions, they could bury it amidst other tags on that graph easily. Whereas you wouldn't be able to do that for any of the other tags because there are so many more views in the first place.

share|improve this answer
    
The scala population is small, I agree. But are you saying that makes the statistics invalid? The sample size is large enough (269), and therefor is statistically relevant and comparable. (colbol, on the other hand, is a mere 30 and arguable too small.) I think the scala skewing is solely because early adopters have fanboi tenancies. –  Stu Thompson Sep 3 '09 at 11:58
    
I'm not saying that it makes the statistics invalid, I'm saying that because the population is small it is easier for a few enthusiastic individuals to wildly affect the ratio. Once the views go up, it will probably fade into the morass with the rest of them –  devinb Sep 3 '09 at 12:12
    
OK, understood. I agree. –  Stu Thompson Sep 3 '09 at 12:15

That big, black dot is terrifying...I mean, I knew subjective was popular, this is is just insane.

share|improve this answer
2  
In all my other graphs, that is where Jon Skeet's bubble normally resides. –  Stu Thompson Sep 3 '09 at 11:30
    
It would be interested to see tag-swarming over time for groups of users. Kinda like "trends." Last week was such a subjective week, following a heavy C# month. –  Jonathan Sampson Sep 3 '09 at 11:32
2  
That's no black dot... –  TheTXI Sep 3 '09 at 11:34
    
define 'groups of users' please –  Stu Thompson Sep 3 '09 at 11:35
    
@Stu It would be more of an explicit group. Almost fanboi-ish. I mean, you could programmatically determine a group of active/popular users, but that's not necessary :) –  Jonathan Sampson Sep 3 '09 at 11:43

You must log in to answer this question.

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