After hearing that homework questions are far too common one too many times, I decided to take the initiative and try to prove whether homework questions are now the primary new content of the site.

I wrote a query in the Data Explorer engine to pull out the number of questions asked each day with and without the tag that are still open (as in, not closed). Then I plotted those counts per day. The results are:

The plot was made in R, the code to repeat is:

data <- read.csv('QueryResults.csv')
data$Date <- as.Date(data$Date, "%Y.%m.%d")
melted <- melt(data[,c('Date','Homework','NotHomework')],id.vars=1)
ggplot(melted, aes(x = Date, y=value, fill=factor(variable))) + geom_bar(stat="identity") + ylab("Number of Questions") + scale_x_date(breaks = seq(min(melted$Date), max(melted$Date), 15)) + theme(axis.text.x = element_text(angle=45, hjust=1, vjust=1))

A summary of the data:

 > summary(data)
      Date               Homework       NotHomework   
  Min.   :2010-08-24   Min.   :0.0000   Min.   : 1.00  
  1st Qu.:2011-05-12   1st Qu.:0.0000   1st Qu.:11.00  
  Median :2011-11-22   Median :1.0000   Median :15.00  
  Mean   :2011-11-22   Mean   :0.9717   Mean   :15.56  
  3rd Qu.:2012-06-03   3rd Qu.:1.0000   3rd Qu.:19.00  
  Max.   :2012-12-15   Max.   :9.0000   Max.   :39.00

So what did I learn?

  1. Homework questions are trending upwards. But so are non-homework questions. So more questions of all types is a good thing.
  2. There are noticeable spikes in the data that seem to correspond to the beginning and ending of semesters, when people need the most help.
  3. Without data to back up statements, it's just wasted time.

Since this is a Q&A site, let me get to my questions:

  1. What do you conclude from this data?
  2. Can we please stop saying that homework questions are swamping out non-homework questions since the data doesn't support it?
Right click and say "View Image" to get a better view of the data if needed. –  tpg2114 Dec 19 '12 at 13:14
Note that the SEDE is updated biweekly (or weekly, forgot which). So the data may not reflect the current status. Also: thanks for this, I was planning to do something similar but I'm no good at data analysis (and I'm too lazy to write sql). I may post a full answer later :) –  Manishearth Dec 19 '12 at 13:20
Yes, but the latest date pulled in the query was 15 Dec 2012, so it was updated recently enough. I don't think the past 4 days are going to reveal anything stunningly different. –  tpg2114 Dec 19 '12 at 13:23
Ah, k then. Anyway, that was just a standard disclaimer-type thingy. Not everyone knows about the SEDE, better to clear out some common misconceptions :) –  Manishearth Dec 19 '12 at 13:27
@tpg2114 look at the level and topics of questions generally before and after the election, independent from the fact if the homework tag has actually been attached. Things have stunningly changed. More theoretical topics are apart from some sporadic popular questions completely absent since then and the few still incoming advanced or more technical questions get no longer answered. John Rennie has noticed it too as you can see in the comments below his question. But of course he is right, –  Dilaton Dec 19 '12 at 16:06
if people interested in more fundamental/theoretical topics who would ask and answer more advanced questions are gone (which seems to be the case), nothing can be done about it. I just have to accept that I am some kind of left behind and nobody else is interested these things anymore. Thats live it seems :-/. –  Dilaton Dec 19 '12 at 16:09
The plot would be much clear if you also plot the ratio of the homework/all questions, and if you group those data in a time period of, say, a week or a month to reduce statistical fluctuation (homework are more weekly based, also daily is not good measure because of the timezone). You should also take a look of the other popular tag, such as quantum-mechanics, which are also grow in the same time period. It will make you conclusion more sound –  hwlau Dec 19 '12 at 17:39
@hwlau Unfortunately I have my actual research work to do right now or I'd play with the data :) But I posted the link to get the raw data, you are free to pull the information and plot it in cooler/better ways and post an answer. –  tpg2114 Dec 19 '12 at 17:58
@tpg2114 I also have some research to do at the moment. So I just leave some idea for the further investigation –  hwlau Dec 19 '12 at 23:08
A physics question is a physics question. Why bother with distinguishing between them? –  Antillar Maximus Dec 23 '12 at 1:25
This just got bumped by community, and in looking at the graph my first thought was that from a data analysis POV it might be helpful to bin by the calendar week---that'll remove the fuzzy appearance and increase the statistical significance of each bin. –  dmckee Apr 7 at 0:33
@dmckee Good point -- I actually saw it bumped, read it, looked through it and thought "Man, who would use R? I bet there's cool stuff we can do with the analysis, I should talk to whoever posted it!" Yeah... totally forgot about this one, and that I knew how to use R at some point. –  tpg2114 Apr 7 at 0:48

1 Answer 1

Well, it seems like you already have the answer: both types of questions are increasing, showing a growth in the site, not one question replacing the other.


You must log in to answer this question.

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