« first day (35 days earlier)      last day (2062 days later) » 

3:21 PM
Just checking I can use the Chat facility in Meta. The list of relevant tags seems to be approaching a steady state, so it may soon be time to try to use it. (But I only used SQL briefly, 30 years ago, therefore I may have to learn it from scratch again.) Am I doing the right thing by simply posting in the chat room named "Data Explorer (SEDE)", or is there a more specific location for discussion of "A statistical question about statistical questions"? (I'm sorry for being such a noob!)
 
3:52 PM
@CalumGilhooley This is a reasonable room for questions about SEDE.
I have to admit that I do not have a good idea how to handle long list of tags - which is what you are after.
Using OR will be probably rather cumbersome - I have tried this for two tags in this query: Questions with tags probability or statistics per week.
The numbers from that query seem comparable to results from on-site search: week 2, week 3.
In case somebody is wondering, the above is related to this question from Mathematcis Meta: A statistical question about statistical questions
 
4:26 PM
@Martin Running that query for 2018 rather than 2016 seems to show a decrease in the frequency of questions tagged 'probability' and/or 'statistics', from some time towards the end of May. Could this perhaps coincide with an increase in the frequency of questions on probability and/or statistics using more specific tags? Or am I misreading the results (perhaps because of my unfamiliarity with SEDE)?
 
Which query?
I should have probably linked the interval form 2016 to 2018: data.stackexchange.com/math/query/893575/…
Looking at a graph, the value seems to be oscillating. Perhaps longer intervals (per month?) are better to check graph for various trends.
 
@Martin Yes, that seems to be the query of yours that I just re-ran, except that I only ran it from the date 2018-01-01.
 
Here is a graph which rene created for single tag:
Aug 28 at 16:11, by rene
It is not bad at all, I wouldn't be so pessimistic about your SQL Skills. Here is one with the EOMonth function so your graph gets dates on the x-axis: https://data.stackexchange.com/math/query/891716/percentage-of-questions-in-a-gi‌​ven-tag-per-month?Tagname=statistics&Date1=2016-01-01&Date2=2018-08-01#graph
 
I see what you mean about the oscillation. It's tricky stuff, this statistics!
 
Here are both tags (probability and statistics) in the sam graph: Percentage of questions in given two tags per month
 
4:49 PM
Although as you say using OR for a long list of tags would be cumbersome, is there any reason to think that it would hit some sort of buffer limit, or in some other way not work? If not, then when my list is complete enough, I could presumably fork one of your queries, inserting a laboriously hand-edited long chain or "OR"s. Would that be daft?
 
That's probably question for somebody more experienced with SEDE.
 
I'll not try it just yet! Thank you for all your help.
 
But my opinion is that using a few big tags (rather than many tags including the smaller ones) could give you a reasonable idea whether there is some increasing or decreasing trend.
After all, it is recommended practice to include a large tag specifying area where the question comes from.
 
Yes, that does seem likely, and it gives me something to do in the meantime, identifying a reasonable short list of tags to use. In practice people don't always seem to use one of those two very big tags, so it is probably worth adding a few more.
 
 
4 hours later…
8:39 PM
@Martin How about this: data.stackexchange.com/math/query/893610/… / cc @CalumGilhooley
That allows you to give in a single field a comma separated list of tags you want to see those percentage for
It is a bit of juggling by generating a dynamic query (using the PIVOT statement fom TSQL) in a variable and then run that sql with the EXEC statement
 
9:07 PM
@rene That's handy. What about a merged version, i.e. a graph or table o the history of the percentage of questions with one or more of the listed tags?
 
If I haven't misunderstood you (I'm having a bad day, so it's possible!), the Results tab shows the table of values corresponding to the graph that is displayed when one clicks the link you posted at 21:39. That's useful information, because I didn't know there was a way to see both the graph and the table. (I'm new to all this.) But I meant, is it possible to tabulate/graph a single percentage, merging the sets for the individual tags?
 
9:33 PM
@CalumGilhooley to the first question the answer is yes.
I think I don't understand what you mean by graph a single percentage, merging the sets for the individual tags. I thought my query did just that. What should the result look like, according to you?
 
9:47 PM
@rene Martin's first query today gave a table of the percentage of questions that had either or both of the tags 'probability', 'statistics'. Would it be possible, similarly, to adapt your query so that it graphs and tabulates only a single number, which is the percentage of questions having one or more of the tags in a given list?
 
ok, I understand. Hold on
 
 
1 hour later…
11:19 PM
@CalumGilhooley it took a while but this query does it per month: data.stackexchange.com/math/query/… for any tags you give it.
Ok, I'm out for the night ....
 

« first day (35 days earlier)      last day (2062 days later) »