Member Contribution Analysis: A Case Study

Discussion in 'About the Members' started by DaveC426913, Oct 27, 2018.

  1. DaveC426913 Valued Senior Member

    In the 'Evidence that God is real' thread, James R separated out what he considers off-topic posts from on-topic posts.

    I find that intriguing. It lends itself to a little bit of analysis, in my opinion. We have a hot topic, with a large dataset of over 1000 posts, but limited to a very manageable number of contributors (33).

    Frankly, this did not break out how I thought it would, and because of that, I've chosen to post it for all. I don't think it disparages anyone in particular, but is fascinating nonetheless.

    Now, take this with a grain of salt: the distinction between on-topic and off-topic was JamesR's alone, so this chart is necessarily a reflection of that.

    All this data is publicly available here:
    (data only covers posts up to the time of separation at 7:48PM, Thurs Oct 26)
    Please forgive any name misspellings.

    I will leave it in the hands of the mods as to whether this thread remains open.

    Please Register or Log in to view the hidden image!

    Please Register or Log in to view the hidden image!

    Write4U likes this.
  2. Guest Guest Advertisement

    to hide all adverts.
  3. DaveC426913 Valued Senior Member

    In my estimation, the highest quality contributor to this thread is ... QuantumQuack. He posted 37 times (almost exactly average) but 100% of his posts were on-topic.

    (With the exception of one meta-post, while wearing a moderator hat, JamesR would have 61 posts, with 100% on-topicality, but I've presumed upon him to recuse himself from qualifying in his own thread.)
  4. Guest Guest Advertisement

    to hide all adverts.
  5. DaveC426913 Valued Senior Member

    Something to note: this is not a very fair analysis, counting only number of posts. Many on-topic posts are pages long. That would skew the results differently. A better (but practically impossible) analysis would use word-count, or at least line count.
  6. Guest Guest Advertisement

    to hide all adverts.
  7. James R Just this guy, you know? Staff Member

    I'd like to say that this thread is entirely DaveC's baby. I'm merely another poster here.

    I expect the inevitable complaint that this is totally biased because I was the person who split off posts to the "off-topic" thread, which is fair enough.

    I also expect the complaint that my judgment about what was and wasn't on-topic for the thread is faulty.

    Take one example: I moved quite a few of Jan Ardena's posts to the off-topic thread because they consisted of little more than repetitive exortations to (and I'm paraphrasing here) "Google William Lane Craig. You'll find my evidence in the search result, somewhere." Jan would no doubt argue that this repetitive nonsense is, in fact, on topic.

    Some posts that were moved were also fallout from Jan's antics. For example, if Jan wrote "But I did provide evidence. Go google William Lane Craig" and somebody else replied "Telling us to search google isn't providing evidence", and that went back and forth 10 or 20 times, then the posts from both parties mostly ended up in the off-topic thread, which might give an unfair impression that the people urging Jan to provide the evidence he claimed to have were posting off-topic. In other words, these kinds of posts, while arguably not off-topic, were moved because they created useless, repetitive clutter in the thread, not advancing the discussion.

    I did leave enough examples of Jan's evasiveness in the thread to leave a record demonstrating his lack of good faith there.
  8. DaveC426913 Valued Senior Member

    Yes. I take full b̶l̶a̶m̶e̶ credit for the data-wrangling and analysis. And I concur that the litmus test of on/off topicality is somewhat arbitrary.

    I could have plotted this out differently (there are lots of options). Instead of percent on-topic as the X axis, I could have just has number on-topic. Then both axes would be open-ended.
    Last edited: Oct 27, 2018
  9. DaveC426913 Valued Senior Member

    Same data, different visualization.

    Shows On-topic posts as a function of Total posts.
    Axes are logarithmic (less weight on absolute count).

    Not all points are identified.
    Opponents of thread thesis (loosely, theist* camp) are in orange.
    Proponents of thread thesis (loosely, atheist* camp) are in blue.
    * only as far as I could identify. Raise your hand if I misidentified you.
    Also, yes, these are binary, confining labels.

    Please Register or Log in to view the hidden image!

  10. Beer w/Straw Transcendental Ignorance! Valued Senior Member

    Well, in my estimation I should have got 100% on the awesomesauce --would have voted for me all the way!

    In contrast, I'm a lot more intrigued on making a personal mechanical pencil and a pen with a 3D printer. I think my first post in the thread was a Brickleberry video...
  11. James R Just this guy, you know? Staff Member

    I think this is a much clearer way to present the data. I should probably try to work out why that is.

    I suppose it's that you can immediately see where posters fall relative to the 50%, 75% and 100% lines.

    I suppose you tried it with a linear scale, too? Does that separate the data points too much, or make those 50%, 75%, 100% lines too close together?

    Out of interest, what software did you use to make the plot?
  12. DaveC426913 Valued Senior Member

    I think so too.

    Hm. Except it's not right.
    Musika's score is 70.2%, yet my graph, above, shows him above the 75% line.
    I suspect that the 75% and 50% lines should not be straight. They should gently flatten out.
    Unfortunately, for reasons that will become apparent as you continue to read, that is easier said than done.

    Using a linear graph just squishes it all into the lower left corner.

    Please Register or Log in to view the hidden image!

    I haven't yet figured out a way of spreading the data points out so you can see them. Logarithmic does that the best.

    That would be Google Spreadsheet and Photoshop.

    Google Spreadsheet does all the data plotting and stuff; I have to apply the 100%, 75% and 50% lines - as well as member labels - in PS.
    Last edited: Oct 28, 2018
  13. DaveC426913 Valued Senior Member

    Hmph. I tried to sign up to MatLab to see if it could render the graphs with percent curves. Couldn't get past the license screen. It simply says I do not have a license.

    Also: I wonder if this thread isn't getting much traffic because of the boring title. Maybe I should change it to something less ... dry.

Share This Page