Not sure if this is the right place to post it and I hope I don't stress you to much with my distributions.
I did some simulations to understand what kind of data we get. Globals are displayed when loot >= 50 PED. From
my previous analysis I've got the exp distribution. There was overestmation in the beginning. With my simulation I could reproduce the same effect, so it's mainly from sampling, and hence no bias.
Unfortunately the exp dist. has some interesting characteristics and we can't distinct if an exp is left truncated or left shifted. Therfore we have 2 possibilities and we can't find the right one with global data.
1) global loot has an exp dist with mean 39-48 and is shifted by 50 PED.
2) global loot has an exp dist with mean 39-48 and is truncated at 50 PED.
If 1) is true, then global loot has probably no influence on loot below 50
PED.
If 2) is true, then global loot has an influence on loot below 50 PED. So pedders like 10, 15 Pedder or silent globals like 40 PED ones, do all belong to the same distribution. Furthermore, if 2) is the case then we see only the last third of all globals and hence the real sample size is 3 times as much.
I have only 350 data points atm, so mean for exp is not estimated very accurately yet, therefore the 39-48 range. I'll need about n=1000 to get a n error range of max 4 PED and n=10000 for 2 PED.