Firstly, how would one obtain such a distribution and its evolution with time?
Secondly, what kind of inferences can be drawn from such a distribution? (I could try this, if I knew answer to previous question.)
$\begingroup$
$\endgroup$
4
-
$\begingroup$ Interesting question, but a bit open-ended (at least its second part). But I would like to know the answer too. $\endgroup$– Piotr MigdalNov 3, 2012 at 11:02
-
2$\begingroup$ The question of what kind of inferences can be drawn from such data strikes me as potentially limitless--I'm sure there are more possible inferences than one would care to count (most of them quite trivial). @PiotrMigdal has given you a good overview of how to get the data; you now need to develop a more focused set of questions. Note, however, that having data can help w/ that: you explore the data, especially (IMO) making plots, think about what it might mean, & repeat. Be sure to retain a hold-out set, though, or you'll never know if your potential insights are worthwhile. $\endgroup$– gung - Reinstate MonicaNov 3, 2012 at 12:26
-
$\begingroup$ @gung thanks for the advice. $\endgroup$– user13107Nov 3, 2012 at 14:33
-
1$\begingroup$ For a graphical example of reputation time series data (and the code to produce it) please see mathematica.stackexchange.com/questions/13705. $\endgroup$– whuber ModNov 6, 2012 at 17:46
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
3
When it comes to getting the data - just go StackExchange Data Explorer and write a SQL query (or find one, or ask me if you have problems with that).
There are other alternatives like parsing StackExchange Data Dump (but requires download a large file and parsing it) or using StackExchange API (but is not that fast, so for such queries it only makes sense for small sites; e.g. beta sites, as they are not covered in neither the Data Explorer nor the Data Dumps).
To some extent SE measures such reputation distributions - see sites at Area51. For fresh sites, it count number of users with rep>=200, rep>=2000 and rep>=3000, just to see if there are both "normal" users (who ask or answer more than one or two questions) and some very dedicated ones.
But for older sites, reputations (as it is cumulative) may be not as informational (unless you cont it in the last 3 months or so).
-
$\begingroup$ I will check the link you mentioned. So, you mean it's not possible to get the 'evolution' part, unless one extracts the info at regular intervals from today onwards? $\endgroup$ Nov 3, 2012 at 14:37
-
1$\begingroup$ @user13107 It is possible, but (AFAIK) not with Data Explorer. Either download consecutive data dumps (each 3 months) or use API - see api.stackexchange.com/docs/full-reputation-history. $\endgroup$ Nov 3, 2012 at 15:23
-