-4

This question is inspired from this: Length limit for question titles that have nothing to do with Unicorns (we don't really need that many characters just for the title, do we? Isn't that what the question body is for? Some lower length limit would help here. Seriously.)

I would like to know the relationship between the activeness of a question and its number of characters in the title. How many characters in the question title will be optimal? I'm curious. If you can visualize it to a plot, then it will be great.

  • Well then.. enjoy :) – ʞunɥdɐpɐɥd Jan 13 '15 at 19:45
  • Well, the raw average length of titles on Stack Overflow is 51: data.stackexchange.com/stackoverflow/query/264686/… – Shadow9 Jan 13 '15 at 20:14
  • Another interesting query, though can't really make any useful conclusion out of it. – Shadow9 Jan 13 '15 at 20:18
  • No logical fallacies here... – user1228 Jan 13 '15 at 20:35
  • How would you define activeness? – rene Jan 13 '15 at 20:41
  • 1
    It is not the length that matters, it is how it is used (the length of the title, that is) – user273376 Jan 14 '15 at 1:01
4

It really depends on how you define activeness but by taking the statistical columns that exist in the post table I'm able to have SEDE produce a graph:

enter image description here

from this SEDE query:

select 
   len(title) as [title len]
 , sum(case when acceptedanswerid is null then 1
       else 0
       end)/800 as accepted
 , avg(viewcount)/200 as [view count]
 , avg(datediff( d, creationdate, nullif(lasteditdate,creationdate)))/10 as [edit duration]
 , avg(datediff( d, creationdate, nullif(lastactivitydate,creationdate)))/15 as [active duration]
 , avg(answercount)*10 as [answer count]
 , avg(commentcount)*10 as [comment count]
from posts p
where p.posttypeid = 1
group by len(title)
order by len(title) desc

This query takes all questions (PostTypeId=1) from the Posts table and sum or averages the statistical values available on each row of a post.

In this case I counted accepted answers, number of views, the duration in days while the post was edited, the duration in days the post was active, the number of answers on the post and the number of comments on the post. On all measures I fiddled with the range to fit them nice together...

What does this tell us? I have really no idea because I had to promise my math teacher to never do someting with statistics for real after my graduation.

I see some kind of distribution (is it Poisson?) for accepted answers...

What strikes me is that for really short or really long titles all statistics vary heavily. That can be due to the title or to other influences to be reasearched.

  • why did you have to promise that? – Ooker Jan 14 '15 at 1:03
  • The graph tells us that the typical question has a title about 43 characters long, and that there is a slight negative correlation between title length and view count. The rest of it is statistical noise due to very few questions having long or short titles. – Mark Jan 14 '15 at 4:58
  • @Ooker because I told him stats are a lie .... – rene Jan 14 '15 at 9:18

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