Show Posts
Time Series on monthly statistics of forum (new users, new topics, new posts)
Plots
1) New members and new topics
A quick overview. It seems that monthly statistics on new members and new topics tend to hit peaks each two-year period.
2) New posts
And that one:1) New members and new topics
2) New posts
As I promised yesterday, I made some quick descriptive analysis today.
Notes:
It's not exact intra-month registered accounts over years due to days on which theymos scrapped data.
However, we can assumed those figures are for intra-month data.
ABSTRACT
(1) 50% of observed months have intra-month registered account range from 2311 to 27790 (the interquartile range);
(2) 50% of observed months have intra-month registered account above or below 10432 (the median);
(3) The intra-month registered account fell 91.6% from its all time high in 2017 at 238400 to 20046; whilst the bitcoin decreased nearly 83.9%.
(4) The demotion on Junior Members in September 2018 probably caused intra-month registered accounts fell considerbaly later months.
(5) The difference in means of before and after periods (cut-off timepoint is Sept. 2018) is -63.6%.
Time-series plot
I don't know where I snapshotted the image, due to lack of thread's author, LOL. Still found it and will post link of the thread later (if I can find it). If you can help, please help me to find the thread.
I actually spent around 15 minutes or more to check back the first 80 pages of Merit & new rank requirements topic. 80 pages made within 3 days after the start of merit system, most of them are pyramid quotes. Honestly, my fingers got hurted due to over-scrolling computer mouse, so I gave up for now. Wholy shit, pyramid quotes!
Scanning pages again, still not found on page #130.
Someone else's thread similiarly describes what the time-series plot above shown. 
As the chart shown, total monthly registered accounts in recent months dropped dramatically to the median and mean of the year 2015.
I believe that it is right time to say "Thanks theymos, merit system works and actually effective control spammers!"
Basic statistics:
- Mean +/- standard deviation: 22200 +/- 34774
- Median (interquartile range): 10432 (2311 - 27790)
- Min - max: 3 - 238433.
Interquartile range (from p25 to p75 in raw result table)
Data interpretation:
- 50% of observed total monthly registered accounts (just assumed the dataset is for exactly monthly data) ranges from 2311 to 27790 (the interquartile range);
- 50% of observed months have total registered accounts above 10432 (the median), whilst the rest 50% of months have figures below 10432.
Data source was dumped by theymos there:
Statistics over years
In both medians and means, the year of 2018 has been the highest year with its median and mean of monthly registered accounts are 66895 and 64782, respectively.
- The median means that 50% of months in 2018 have total registered accounts above 66895 or below 66895.
- The interquartile range for 2018 is 38393 to 88373. It means 50% of months in 2018 have figures in range from 38393 to 88373.
I always use median and interquartile range for descriptive statistics because they automatically exclude effects of outliers data points.
Percent changes of intra-month registered account:
The figures decreased 91.6% when bitcoin fell 83.9%.
It gives us a potential causal inference that implementation of merit system and new rank requirements have bigger effects than the falls of bitcoin.
One more finding, the intra-month figures after September of 2018 (Junior Member demotions) fell significantly than the before period. I think it make sense.
Detail calculation presented below.
The means of before and after Sept. 2018 are 80179 and 29190 (respectively)
The percent of difference is 63.6%
Notes:
It's not exact intra-month registered accounts over years due to days on which theymos scrapped data.
However, we can assumed those figures are for intra-month data.
ABSTRACT
(1) 50% of observed months have intra-month registered account range from 2311 to 27790 (the interquartile range);
(2) 50% of observed months have intra-month registered account above or below 10432 (the median);
(3) The intra-month registered account fell 91.6% from its all time high in 2017 at 238400 to 20046; whilst the bitcoin decreased nearly 83.9%.
(4) The demotion on Junior Members in September 2018 probably caused intra-month registered accounts fell considerbaly later months.
(5) The difference in means of before and after periods (cut-off timepoint is Sept. 2018) is -63.6%.
Time-series plot
I don't know where I snapshotted the image, due to lack of thread's author, LOL. Still found it and will post link of the thread later (if I can find it). If you can help, please help me to find the thread.
I actually spent around 15 minutes or more to check back the first 80 pages of Merit & new rank requirements topic. 80 pages made within 3 days after the start of merit system, most of them are pyramid quotes. Honestly, my fingers got hurted due to over-scrolling computer mouse, so I gave up for now. Wholy shit, pyramid quotes!
Scanning pages again, still not found on page #130.
As the chart shown, total monthly registered accounts in recent months dropped dramatically to the median and mean of the year 2015.
I believe that it is right time to say "Thanks theymos, merit system works and actually effective control spammers!"
Basic statistics:
- Mean +/- standard deviation: 22200 +/- 34774
- Median (interquartile range): 10432 (2311 - 27790)
- Min - max: 3 - 238433.
Interquartile range (from p25 to p75 in raw result table)
Data interpretation:
- 50% of observed total monthly registered accounts (just assumed the dataset is for exactly monthly data) ranges from 2311 to 27790 (the interquartile range);
- 50% of observed months have total registered accounts above 10432 (the median), whilst the rest 50% of months have figures below 10432.
Code:
. tabstat regacc, s(n mean sd p50 p25 p75 min max) c(s)
variable | N mean sd p50 p25 p75 min max
-------------+--------------------------------------------------------------------------------
regacc | 112 22199.58 34773.29 10432 2310.5 27790 3 238433
----------------------------------------------------------------------------------------------
Data source was dumped by theymos there:
We're at a local minimum, but registrations jump around a lot:
Code:
+---------------------+---------------+
| 30day starting | registrations |
+---------------------+---------------+
| 2009-11-02 00:00:00 | 3 |
| 2009-12-02 00:00:00 | 8 |
| 2010-01-01 00:00:00 | 8 |
| 2010-01-31 00:00:00 | 22 |
| 2010-03-02 00:00:00 | 5 |
| 2010-04-01 00:00:00 | 22 |
| 2010-05-01 00:00:00 | 15 |
| 2010-05-31 00:00:00 | 33 |
| 2010-06-30 00:00:00 | 325 |
| 2010-07-30 00:00:00 | 159 |
| 2010-08-29 00:00:00 | 148 |
| 2010-09-28 00:00:00 | 190 |
| 2010-10-28 00:00:00 | 204 |
| 2010-11-27 00:00:00 | 383 |
| 2010-12-27 00:00:00 | 384 |
| 2011-01-26 00:00:00 | 938 |
| 2011-02-25 00:00:00 | 2202 |
| 2011-03-27 00:00:00 | 2737 |
| 2011-04-26 00:00:00 | 4936 |
| 2011-05-26 00:00:00 | 14048 |
| 2011-06-25 00:00:00 | 6250 |
| 2011-07-25 00:00:00 | 3903 |
| 2011-08-24 00:00:00 | 2938 |
| 2011-09-23 00:00:00 | 1983 |
| 2011-10-23 00:00:00 | 1965 |
| 2011-11-22 00:00:00 | 1835 |
| 2011-12-22 00:00:00 | 1812 |
| 2012-01-21 00:00:00 | 2082 |
| 2012-02-20 00:00:00 | 1901 |
| 2012-03-21 00:00:00 | 2158 |
| 2012-04-20 00:00:00 | 1969 |
| 2012-05-20 00:00:00 | 2057 |
| 2012-06-19 00:00:00 | 2040 |
| 2012-07-19 00:00:00 | 2145 |
| 2012-08-18 00:00:00 | 2419 |
| 2012-09-17 00:00:00 | 2661 |
| 2012-10-17 00:00:00 | 2958 |
| 2012-11-16 00:00:00 | 2763 |
| 2012-12-16 00:00:00 | 2611 |
| 2013-01-15 00:00:00 | 3320 |
| 2013-02-14 00:00:00 | 4933 |
| 2013-03-16 00:00:00 | 13816 |
| 2013-04-15 00:00:00 | 18162 |
| 2013-05-15 00:00:00 | 11318 |
| 2013-06-14 00:00:00 | 8095 |
| 2013-07-14 00:00:00 | 6703 |
| 2013-08-13 00:00:00 | 5505 |
| 2013-09-12 00:00:00 | 4137 |
| 2013-10-12 00:00:00 | 5632 |
| 2013-11-11 00:00:00 | 26804 |
| 2013-12-11 00:00:00 | 28776 |
| 2014-01-10 00:00:00 | 29576 |
| 2014-02-09 00:00:00 | 29839 |
| 2014-03-11 00:00:00 | 22534 |
| 2014-04-10 00:00:00 | 18065 |
| 2014-05-10 00:00:00 | 13071 |
| 2014-06-09 00:00:00 | 12180 |
| 2014-07-09 00:00:00 | 10561 |
| 2014-08-08 00:00:00 | 10925 |
| 2014-09-07 00:00:00 | 9418 |
| 2014-10-07 00:00:00 | 8218 |
| 2014-11-06 00:00:00 | 7974 |
| 2014-12-06 00:00:00 | 8136 |
| 2015-01-05 00:00:00 | 23256 |
| 2015-02-04 00:00:00 | 36913 |
| 2015-03-06 00:00:00 | 33500 |
| 2015-04-05 00:00:00 | 8450 |
| 2015-05-05 00:00:00 | 8577 |
| 2015-06-04 00:00:00 | 10303 |
| 2015-07-04 00:00:00 | 11773 |
| 2015-08-03 00:00:00 | 9725 |
| 2015-09-02 00:00:00 | 7644 |
| 2015-10-02 00:00:00 | 68601 |
| 2015-11-01 00:00:00 | 44059 |
| 2015-12-01 00:00:00 | 39244 |
| 2015-12-31 00:00:00 | 37106 |
| 2016-01-30 00:00:00 | 42233 |
| 2016-02-29 00:00:00 | 24213 |
| 2016-03-30 00:00:00 | 19336 |
| 2016-04-29 00:00:00 | 16031 |
| 2016-05-29 00:00:00 | 19935 |
| 2016-06-28 00:00:00 | 12582 |
| 2016-07-28 00:00:00 | 10646 |
| 2016-08-27 00:00:00 | 9986 |
| 2016-09-26 00:00:00 | 9844 |
| 2016-10-26 00:00:00 | 10718 |
| 2016-11-25 00:00:00 | 10691 |
| 2016-12-25 00:00:00 | 12429 |
| 2017-01-24 00:00:00 | 13664 |
| 2017-02-23 00:00:00 | 13893 |
| 2017-03-25 00:00:00 | 15276 |
| 2017-04-24 00:00:00 | 20585 |
| 2017-05-24 00:00:00 | 31381 |
| 2017-06-23 00:00:00 | 35866 |
| 2017-07-23 00:00:00 | 40726 |
| 2017-08-22 00:00:00 | 45026 |
| 2017-09-21 00:00:00 | 70477 |
| 2017-10-21 00:00:00 | 95568 |
| 2017-11-20 00:00:00 | 144762 |
| 2017-12-20 00:00:00 | 238433 |
| 2018-01-19 00:00:00 | 136872 |
| 2018-02-18 00:00:00 | 107237 |
| 2018-03-20 00:00:00 | 87087 |
| 2018-04-19 00:00:00 | 89659 |
| 2018-05-19 00:00:00 | 83999 |
| 2018-06-18 00:00:00 | 72430 |
| 2018-07-18 00:00:00 | 61360 |
| 2018-08-17 00:00:00 | 43451 |
| 2018-09-16 00:00:00 | 43516 |
| 2018-10-16 00:00:00 | 35335 |
| 2018-11-15 00:00:00 | 31502 |
| 2018-12-15 00:00:00 | 20733 |
| 2019-01-14 00:00:00 | 20046 |
| 2019-02-13 00:00:00 | 2763 | [PARTIAL PERIOD]
+---------------------+---------------+
Statistics over years
In both medians and means, the year of 2018 has been the highest year with its median and mean of monthly registered accounts are 66895 and 64782, respectively.
- The median means that 50% of months in 2018 have total registered accounts above 66895 or below 66895.
- The interquartile range for 2018 is 38393 to 88373. It means 50% of months in 2018 have figures in range from 38393 to 88373.
I always use median and interquartile range for descriptive statistics because they automatically exclude effects of outliers data points.
Code:
. tabstat regacc year10 year11 year12 year13 year14 year15 year16 year17 year18, ///
> s(n mean sd p50 p25 p75 min max) c(s)
variable | N mean sd p50 p25 p75 min max
-------------+--------------------------------------------------------------------------------
regacc | 112 22199.58 34773.29 10432 2310.5 27790 3 238433
year10 | 13 146 144.3185 148 22 204 5 384
year11 | 12 3795.583 3560.192 2469.5 1900 4419.5 938 14048
year12 | 12 2313.667 353.35 2151.5 2048.5 2636 1901 2958
year13 | 12 11433.42 8798.91 7399 5219 15989 3320 28776
year14 | 12 15041.42 8092.689 11552.5 8818 20299.5 7974 29839
year15 | 12 24992.25 19299.52 17514.5 9151 37009.5 7644 68601
year16 | 12 16553.67 9350.784 12505.5 10668.5 19635.5 9844 42233
year17 | 12 63804.75 67452.29 38296 17930.5 83022.5 13664 238433
year18 | 12 67431.75 34995.66 66895 38393 88373 20733 136872
----------------------------------------------------------------------------------------------
Percent changes of intra-month registered account:
The figures decreased 91.6% when bitcoin fell 83.9%.
It gives us a potential causal inference that implementation of merit system and new rank requirements have bigger effects than the falls of bitcoin.
One more finding, the intra-month figures after September of 2018 (Junior Member demotions) fell significantly than the before period. I think it make sense.
Detail calculation presented below.
Code:
* BTC price
. di (19800-3190)*100/19800
83.888889
* Intra-month registered accounts
. di (238400-20046)*100/238400
91.591443
The means of before and after Sept. 2018 are 80179 and 29190 (respectively)
The percent of difference is 63.6%
Code:
. tabstat before after, s(n mean sd p50 p25 p75 min max) c(s)
variable | N mean sd p50 p25 p75 min max
-------------+--------------------------------------------------------------------------------
before | 9 80179 30633.52 83999 61360 89659 41451 136872
after | 3 29190 7570.575 31502 20733 35335 20733 35335
----------------------------------------------------------------------------------------------
Code:
. di (80179-29190)*100/80179
63.593959
