1. Please take 30 seconds to register your free account to remove most ads, post topics, make friends, earn reward points at our store, and more!  

Thread Count: TB Classifieds

Discussion in 'Miscellaneous [BG]' started by bassRunner, Jun 4, 2014.


  1. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    Hey all. Here's my second project in analyzing some of the data in the talk bass classifieds. You can find my first project here: http://www.talkbass.com/threads/price-distribution-tb-classifieds.1074921/

    Anyway, what I've done here is simply count the number of threads started each month in the for-sale-bass-guitar forum. It looks like September 2005 was the first month with more than a couple of threads. Do tb veterans remember when the classified forum went live?

    Honestly, I had higher hopes for this data. Of course there are several other forums (amps, effects, parts, etc.) to search that may be better for teasing out the information I was hoping to find. Mainly, I was hoping to be able to identify (by eye) seasonal variations. I was thinking that November and December would be months with high thread counts as people tend to have lots of expenses around the holiday season. I also thought tax season might make itself obvious. We can see some clear ups and downs from month, but nothing jumps out at me saying "holiday season."

    I was also thinking the economic downturn in 2008 would be pretty obvious, but it looks like there is a more or less linear increase in threads per month beginning in 2005 (well before the downturn) that extends to about 2010, after which things level off.

    Just a note: I have accounted for the difference in number of days in each month. All things equal, March will naturally have about 10% more threads started than February. I've removed this effect in my code.

    I'll run my data collection code on the other classified forums and post later. For now, enough from me, here's the data! Any ideas?

    thread_count_for_sale_bass_guitars.
     
  2. Stewie26

    Stewie26 Supporting Member

    It would be cool if you had another line behind the blue one with the TB membership growth.
     
  3. IGotGas

    IGotGas Cajun Rocker

    Sep 26, 2011
    Baton Rouge, LA
    Data! Data! Data! I live for data!

    So, yes, membership growth could be a great insight. Also, if there were a way, showing not only listings by month, but solds as well (harder to grab and verify that data). As an aside, days to sell (and of course all us data nerds would want to break that down by make, model, strings, finish and if it's actually THE BEST bass for metal. Or not).
     
  4. Can you add a line of classifieds growth after I joined this forum haha.
     
  5. R&B

    R&B Both kinds of music: Rhythm AND Blues! Supporting Member

    It's steep! :p

    Digging this data analysis. Can you put up a bar chart by month? You could superimpose the available years. Maybe we'll figure out the best time to sell the best bass for metal.
     
    bassRunner likes this.
  6. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    Okay, this was such a good idea, I did it on my lunch break. This only includes data from 2006 through 2013 because 2005 and 2014 do not have data for all 12 months. No surprise, November has the highest thread count. However, I think there's an argument to be made that the month to month differences aren't significant enough to make it worth delaying your for sale thread.

    months_only_for_sale_bass_guitars.
     
    R&B likes this.
  7. R&B

    R&B Both kinds of music: Rhythm AND Blues! Supporting Member

    OK now, during your afternoon break, let's see separate bars for the same month in each of the years (on the same x-axis). The y-axis scale will have to come down by a factor of ~7 to display well. :) Then, error bars with standard deviations. Oh wait, this is TB!! :cool:
     
    SunnBass and bassRunner like this.
  8. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    Ah, but the error estimates would be skewed by the 2006 data which is much lower, possibly because membership was lower back then as some have suggested. Better to bootstrap the data (rofl, this is going to far). And yes, this is TB, not my thesis. :)
     
  9. Trendspotting: More and more people are discovering that the bass is not "easier than guitar because it only has four strings," quitting in frustration and selling their gear to us!
     
  10. Lobster11

    Lobster11 Supporting Member Supporting Member

    Apr 22, 2006
    Williamsburg, VA
    I was puzzling over the fact that your data, as you noted, don't seem to show any obvious effect of the economic downturn in 2008, and I came up with this hypothesis: Although there might well have been an increase in the number of TBers selling basses at this time because they needed the dough -- i.e., because they lost a job or had other financial problems -- there might have been a corresponding decrease in the number of folks selling old basses because they intended to replace them with new or better ones. I'll bet that if you looked at the reasons people gave for selling -- which often (but not always) are stated in the OP -- you'd be able to see that difference. Alternatively -- and this would be easier to quantify -- you could look to see if each OP did or did not express interest in a trade, since people selling for financial reasons would presumably be less interested in trades than those looking to change or upgrade.

    Of course, another possibility is that any effect of the recession might be hidden by confounding effects of membership increases. I have no idea, but maybe it's the case that around 2008 or so there was a huge increase in TB membership, such that proportionally the increase in bass-sale threads looks (as expected) like a slowdown. If you could get your hands on monthly TB membership totals for the last eight years, you could compute some kind of per-capita variable, and perhaps that variable would show a more noticeable effect.

    P.S. This is a good thread in which to use words like "hypothesis" and "confounding," right?
     
    R&B and bassRunner like this.
  11. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    I contacted the admins about getting the data for the number of members that joined talkbass each month since it began. I'm not aware that we as users have access to any master list of members where I could easily read the "Joined:" dates, but the admins surely would have access to such a list. We'll see what they say, I'm sure they're incredibly busy, but I think this data would add a lot to our discussion. Thanks for the idea!
     
  12. Cool to see. Thanx
     
  13. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    Here's the data for the amps classified forum. The data exhibit the same broad behavior, linear increase until about 2010 after which it levels off. There are considerably fewer threads in this forum: about 1000 pages with 30 threads per page as opposed to the 1700 pages in the bass guitar classified forum.

    thread_count_for_sale_amps_preamps_and_cabinets.
     
  14. paul

    paul Staff Member Founder Administrator

    Jul 20, 2000
    Texas
    The number of registrations/month is a pretty useless stat, since in the early days we had no bot protection. We had at one point maybe 70-100 bots registering per day. Over the past few years we had some protection, but it didn't stop them all. Now we have great protection against bots. So you'd see a steady decline in registrations/month that has nothing to do with activity here.
     
    bassRunner likes this.
  15. bassRunner

    bassRunner

    Aug 10, 2012
    Urbana, IL
    I had not considered that bots would screw up membership counts data, which is too bad. I'm very green when it comes to web statistics, and this is a valuable lesson for me. Anyway, thanks, Paul, for taking the time to provide your input, and also thanks for creating such an awesome environment for all of us to Talk Bass! :bassist:
     
  16. Gravedigger Dav

    Gravedigger Dav Supporting Member

    Mar 13, 2014
    Fort Worth, Texas
    Easy - Number of posts / number of members.
     
  17. Lobster11

    Lobster11 Supporting Member Supporting Member

    Apr 22, 2006
    Williamsburg, VA
    I suggested that two posts after the one you quoted, but Paul pointed out this morning (a couple of posts before yours) why membership numbers are probably not reliable until very recently.
     
  18. Gravedigger Dav

    Gravedigger Dav Supporting Member

    Mar 13, 2014
    Fort Worth, Texas
    Yes, I saw that after I posted and I would agree with it. But, doing analysis, you don't always have the best numbers, so you go with what you have knowing it is not exact, but closer than before.
     
  19. Lobster11

    Lobster11 Supporting Member Supporting Member

    Apr 22, 2006
    Williamsburg, VA
    Agreed -- if you are confident that what you have really is "closer than before." If it seems safe to assume that the proportion of "members" that are really bots is relatively stable across time (until new procedures were introduced recently to exclude them), then there's no problem: If the degree to which membership numbers are artificially inflated is essentially held constant across time, the absolute membership numbers would be off but their relative values across time would be valid. On the other hand, if bot-registrations tend to come and go in waves, or otherwise rise and fall dramatically from month to month or year to year, "correcting" for membership numbers might make matters worse. I just don't know enough about Bot World to hazard a guess as to which is more likely to be true.
     
  20. Gravedigger Dav

    Gravedigger Dav Supporting Member

    Mar 13, 2014
    Fort Worth, Texas
    Good points. I do think that since the existance of bots is known, then there must be some available data regarding their numbers over time. I would be hard pressed to believe those who have to deal with bots don't have any idea of their numbers or percentages of actual registrations. If there is knowledge they exist but the values are not known, then there need to be some way to factor their effect or any derived data will be suspect.
    IMHO, if you know your data is flawed, you should share it anyway with disclaimers and notes. It may be someone else has information you don't and can fill in the gaps.
     

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.