Where did you get it?
Can you give me full dataset of the merit system, started in January this year, please.
I need supplementary data for weeks before the 5th or March.
Thanks for all your potential help in advance.
On the 1st week, we had the all-time high, with a spike of 9587 merits! The cause of this is
Enhanced newbie restrictions & requirements when a lot of Jr. Members with no earned merit were de-ranked to newbie status and to rank up again they needed 1 merit.
Nice observation!
After that, we went down week after week till on the week from 09-10-2018 to 04/11/2018 with only 3321 very close to 3186 on the week from 27/08/2018 to 02/09/2018.
Nevertheless, the latest drop of weekly distributed merit has still been in normal range.
I will pay more attention on weeks which have total distributed mertis around 2.5k. Those sorts of low distributed merits are highly potential outliers. Therefore, if they occur, they should be taken into deeply consideration and further forensic analysis in order to discover which sort of factors result in the dramatic drops.
In general, it's hard to see weeks which have statistics are potential outliers (above or below). By now, we have seen only one week which has been potential outliers (above one) at 9684 [maybe more if accounted for early weeks since January]; there has been no observed potential below outlier [based on all weekly data since 5th of March].
In a nutshell, weeks that have total distributed merits above 6.2k or below 2.5k should be taken into consideration because they are highly potential outliers. It is probably that the sorts of outlier-weeks occur due to significant changes in the forum rules/ systems, etc.
Of course, the exact value of outlier thresholds, both above and below, will change over time, but it will be somewhere around those points (2.5k and 6.2k).
You can visit the post for more details
Weekly update on merit distributions since the 5th of March.
Medians should be used as a gold standard figure to measure the level of weekly change. I meant we should observe the change around the medians, which also called as true means.
In my opinion, the median, which fluctuated around 4.4k, should be use as a gold figure to measure changes of weekly distributed merits. Medians are recognised as nearly true means, more closely to true means, and should be used to have more exact overview.
