Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Aklyon]

I woudn't be surprised if a significant percent of recent bans are ameripol-related.

[My Name is Immaterial]

I woudn't be surprised if a significant percent of recent bans are ameripol-related.

Xeron      November 9, 2016      Ban      Trolling
Codician      November 9, 2016      Ban      Trolling
Vendayn      November 11, 2016      Ban      Trolling

I'm going to go with yes, they are.

[TheBiggerFish]

Also BFEL.

[Elephant Parade]

And mainiac, too.

[My Name is Immaterial]

Oh jeez, I missed a few months.
General Update:

No major news from this front. An unexpected spike in new topics on the 27th of March, along with an odd increase in online members around that time. Any ideas as to what that could be? Did the forums get hit by spammers?


In other news, I took a class on how to run statistical regressions and model building, so:
I'm proud to reveal the first model of B12 Posting Trends!

Based on a regression of daily post counts from 1/1/2015 to 12/31/2016, the equation determining how many posts are expected on a given day is:

Posts =  2506.71 + 2.18*Days since last update + -283.18*February + -576.60*March + -615.83*April + -676.63*May + -762.90*June + -928.52*July + -1077.16*August + -1415.67*September +  -1341.61*October + -1552.25*November + -1115.57*December + 120.98*Monday + -101.43*Saturday + -11.37*Sunday + 65.69*Thursday + 121.15*Tuesday + 96.09*Wednesday
R² = .5074
All variable except Sunday and Thursday are statistically significant at the 95% level.

So what the actual fuck did you just read?
I've created a statistical model to determine what factors affected the number of posts on a given day. I looked at a lot of factors, including ban records, update history, day of the week, and a dozen other variables that the math decided wasn't a significant factor. I settled on three variables that are able to predicted 50.74% of the variation in this data set: 1. Days since the last update. 2. Month of the year. 3. Day of the week.
The second two are broken up into what are called fixed effects, or binary variables: it is either Monday (Monday=1), or it is not (Monday=0). This is why there are 17 random variables sitting around; each one is 1 or 0 depending on the day or month you are trying to look at. You may also notice that January and Friday are omitted. This is because Stata, the program I am using, chooses a variable from the set of fixed variables, and sets all the other variables to be compared to that one. Basically, if you're looking at a Friday in January, just look at the constant term, and ignore the month/week variables.
THIS DOES NOT PREDICTED THE RIGHT NUMBER WITH 50% ACCURACY!!! R² is a term used to determine how accurately a regression explains the variance of a data set. If I add a very significant variable that explains a lot of the variation, R² will increase sharply. This model isn't a great one, because there are so many variables that I can't actually measure and include.


I also came up with a model for Monthly totals, but it's not great, and has poor predictive powers, so I'm not going to be talking about it until I've come up with a better one.

Spoiler: Stat outputs for NEEEEEEEEEEEEEEERDS (click to show/hide)

Code: [Select]

xi: reg posts days i.month i.week
i.month           _Imonth_1-12        (naturally coded; _Imonth_1 omitted)
i.week            _Iweek_1-7          (_Iweek_1 for week==Friday omitted)

      Source |       SS           df       MS      Number of obs   =       731
-------------+----------------------------------   F(18, 712)      =     40.74
       Model |  96198825.3        18  5344379.19   Prob > F        =    0.0000
    Residual |  93397910.4       712  131176.841   R-squared       =    0.5074
-------------+----------------------------------   Adj R-squared   =    0.4949
       Total |   189596736       730  259721.556   Root MSE        =    362.18

------------------------------------------------------------------------------
       posts |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        days |   2.177757   .1950183    11.17   0.000     1.794877    2.560637
   _Imonth_2 |  -283.1795   66.53974    -4.26   0.000    -413.8171    -152.542
   _Imonth_3 |  -576.6013   65.56843    -8.79   0.000    -705.3319   -447.8708
   _Imonth_4 |  -615.8251   67.04985    -9.18   0.000    -747.4642    -484.186
   _Imonth_5 |  -676.6282   66.33213   -10.20   0.000    -806.8582   -546.3982
   _Imonth_6 |  -762.9003    67.2682   -11.34   0.000     -894.968   -630.8325
   _Imonth_7 |  -928.5204   67.41095   -13.77   0.000    -1060.868   -796.1724
   _Imonth_8 |  -1077.156    69.2055   -15.56   0.000    -1213.027   -941.2848
   _Imonth_9 |  -1415.672   71.93065   -19.68   0.000    -1556.893    -1274.45
  _Imonth_10 |  -1341.613   74.06899   -18.11   0.000    -1487.033   -1196.193
  _Imonth_11 |  -1552.247   77.57988   -20.01   0.000     -1704.56   -1399.934
  _Imonth_12 |  -1115.574   66.61788   -16.75   0.000    -1246.365   -984.7826
    _Iweek_2 |   120.9837   50.14381     2.41   0.016     22.53629    219.4311
    _Iweek_3 |  -101.4252   49.99382    -2.03   0.043    -199.5781   -3.272237
    _Iweek_4 |  -11.36799   50.12367    -0.23   0.821    -109.7759    87.03989
    _Iweek_5 |   65.69134   49.99832     1.31   0.189    -32.47042    163.8531
    _Iweek_6 |   121.1508   50.14273     2.42   0.016     22.70554    219.5961
    _Iweek_7 |   96.08893   50.13079     1.92   0.056    -2.332919    194.5108
       _cons |    2506.71   56.12025    44.67   0.000     2396.529    2616.891
------------------------------------------------------------------------------

If you want to learn more, shoot me a PM, and I can walk you through it. Alternatively, you can check out the Excel sheets, Sandboxes 5 & 6, where I setup the models. I can also send over the raw Stata files if you want to take a look at them.

[TheBiggerFish]

Yes, there was a spambot in GD on that date.

[My Name is Immaterial]

I guess that answers my unasked about whether or not deleted posts count towards the total. It seems that they do.

[Elephant Parade]

Yes, there was a spambot in GD on that date.

I guess that answers my unasked about whether or not deleted posts count towards the total. It seems that they do.

Has the Mystery of December Eighth finally been solved?!

[Inarius]

The general trend is to have fewer and fewer posts, so I'm surprised that the "days" are with a positive coef, but I hadn't taken into account the "since last release" thing. I'm not sure if you could have a better rē with others coef. Good job for the "day-related" thing, anyway ! That's smart.

[My Name is Immaterial]

Updated!

Updates on my side: my efforts to learn how to scrape the stats page with Python have been thwarted by pip refusing to update/install modules. Please send help.