HQ king triggers

[Bumbeen]

Nope , especially with different Treasure Hunter procs and different NM mixed. It doesn't excude 25% for HQ king so far. 10% or 30% is a lot different and 20 sample size cannot distingish this.

No 20 samples cannot distinguish between 10% or 30%, but that still gives you a good idea of what the drop rate is. As in it's less than 30%, at least. 3/22 in fact is more like +-14% so sub 28% drop rate is almost certain. I don't see how that isn't useful information.

[wish12oz]

No 20 samples cannot distinguish between 10% or 30%, but that still gives you a good idea of what the drop rate is. As in it's less than 30%, at least. 3/22 in fact is more like +-14% so sub 28% drop rate is almost certain. I don't see how that isn't useful information.

Especially considering we have much more then a sample size of 20? I saw 189.

[Nynja]

189 isnt enoug, it must be 1000!!!!!!

[Byrth]

Sorry, I basically read the first few posts and wrote that. Typically the value you multiply the assumed standard deviation by scales with sample size (so lower N or degrees of freedom = higher number). I used 1.96, which is the generic "Hey, you're above 30!" number. There are also some tests that people simply say are not valid until you have >30 sample size.

I've personally not had much issues with that equation I'm using giving very inaccurate results at low N (apart from where p is very close to 100% or 0%). It gave a huge window (almost 1/3 of the probability space), which is appropriate for a low N.

[Nynja]

http://americanresearchgroup.com/moe.html

I used that, presuming Population Size would be the amount of "accepted data" to tell whats accurate, I used 1000, and threw in our data of 189 samples, which gave a margin of error of 6.42%? Did I do that right?

Well, throwing around random population size numbers didnt change much, its always around the 5-7% margin of error range.

[Byrth]

For a two-state event (it happens or it doesn't), I think the binomial distribution error thing is best.
X = how often something happened
N = sample size

X/N +/- (1.96*SQRT( (1 - X/N) * X / N^2) )

or, when p is the ratio:
p +/- (1.96*SQRT( (1-p)*p/N) )

(I would just do it for you, but I can't find the 189 you're talking about)

[Nynja]

Updated loot counts, separated by NM, all presumed with TH.

Adamantoise:
Me 2/6, Gwynplaine 5/10, Karb 0/3, Insaniac 0/4
7/23

Fafnir:
Me 0/3 BG 0/3, Rak 1/8, Karb 1/3, Wiki 1/32, Insaniac 0/11 Gwyn 0/7 Wish12fraand 1/17 Wish12 0/12
4/96

Behemoth:
Me 0/4, BG 0/3, Rak 1/8, Karb 0/4, Saca 0/8, Wiki 3/26, Insaniac 1/9 Wish12 0/8
5/70

thats the 189

[Byrth]

I get 8.5% +/- 4%, so likely 10%. This assumes all three populations have the same probability.

[Karbuncle]

The real problem is that "10%" is with TH8~+ on most all of those kills.

Thats some pretty shitty Drop rate D: Would anyone complain if they upped it to ~20% >_>?

[Raksha]

I get 8.5% +/- 4%, so likely 10%. This assumes all three populations have the same probability.

So 990 k.seals on average to pop HQ king?

That d.ring is gonna take a while >_<.