< ... >
Stuffs that I would like to pinpoint are:
(1) Medians should be used to present almost true means of statistics, such as merits, activities, postcounts, trust points, etc, rather than means.
Because Means significantly affected by outliers.
(2) Percentage change for each week should be compared between total distributed merits during each week with median of weekly distributed merit.
By now, @coinlocket$ compared two weeks in-a-row. This approach has limitations.
For instance, let's take a look at the following image.
For the week started at 24/9/2018, the percentage change compared with previous week that has 9684 merit distributed in total, is -53,4% (as @coinlocket$’s table shows).
Nevertheless, if we use the median for comparison, which sort of result we can have.
Let’s see:
The median till the end of the week started at 24/9/2018 is 4440 (the first row in my table).
The difference between the week and the median is: 4510 – 4440 = 70 merits
Percentage change is: 70*100/4440 =
1.6%.Once again, what does it mean?
It clearly presents that the effect of new rank requirements (aims at Junior Member rank) almost tailed off only after two week since the launch day
(1.6% higher than median is very low effect).Based on the given figure by @coinlocket$, readers can easily get wrong overview that the week had significant changes due to something unknown at that moment; but in reality everything probably come back to normal (with my analytic approach).
(3) For trend analysis, I will do it with time series analysis.
Let's wait and see when we will witness a first statistical significant trend break which might occur due to dramatic changes in the forum.
events such as initial airdrop being spent heavily over the first weeks, and the change of rules for Jr. Member achievement.
By now, there is no trend-break, as we all can see that the change of rules for Junior Member rank has had only temporary effects, which has not lasted too long. Consequently, these sort of temporary effects has not long enough to make a trend break.