I am not sure, but the following findings maybe partially due to effects of the latest restrictions and tightening requirements on Junior member rank.
During the recent 10 weeks, after the start of new tank requirements on Junior members, the weekly merit per post has been on downwards. The figures of peak is 80% higher than the bottom during the period
Of course, there is not good causal-inferance here due to lack of more potential bias variables that have not included, such as:
- Merit per ranks;
- Posts per ranks;
- Decreases of new bounties;
- And more bias factors.
In addition, it is un-clear, but it seems that merit per post tend to increase each four weeks.
It is too soon to go to conclusion that there is monthly periodic because only 10 weeks followed in the dataset.
We need more weekly data points to have a clearer picture.
I would like to disclose one more thing, merit per post during the last 10 weeks.
My approach:1) Total weekly merits: Calculated from real data.
2) Total weekly posts (weekposts): Calculated from real data, except the week #2018w37, which is only the assumed median/ mean from the continuous three weeks (2018w35, 2018w36, and 2018w37). I took it from the OP.
3) Merit per post (meritperpost): = round(merit/weekposts,0.001).
Real dataset is here:
. list, abb(30)
+-------------------------------------------------+
| week merit weekposts id meritperpost |
|-------------------------------------------------|
1. | 2018w37 5630 353813 34 .016 |
2. | 2018w38 7825 286408 35 .027 |
3. | 2018w39 4388 280503 36 .016 |
4. | 2018w40 4271 259694 37 .016 |
5. | 2018w41 3800 263685 38 .014 |
|-------------------------------------------------|
6. | 2018w42 4821 240339 39 .02 |
7. | 2018w43 3945 240045 40 .016 |
8. | 2018w44 3339 219689 41 .015 |
9. | 2018w45 4513 219774 42 .021 |
10. | 2018w46 1740 215135 43 .008 |
+-------------------------------------------------+
Results:Merit per post has fluctuated during the 10-week period, from 0.015 to 0.207 merit per post.
The shocking figure found here is that there is 80-percentage difference between the highest and the lowest compared to the lowest.
. di (0.027-0.015)*100/0.015
80
The last week (#2018w46) is incomplete one, with only half of the week collected in the dataset.
Here is the plot of the ten-week-period merit per post:
Notes:
It is better if having dataset on different ranks in the forum (for merits, posts), but I unfortunately have not had them so far.
