It seems there is a significant change in Charm rate when wearing a Mask of Granite.

 

Typically, according to the 150k+ log on RS Wiki, the rate is as follows: 74% No Charm, 9-10% Gold, 5% Green, 5% Red, and 7% Blue.

 

Wearing a Mask of Granite for a task of 198 Gargoyles yielded the following:

 

68.7% No Charm (136 kills)

 

12.6% Gold Charm (25 kills)

 

11.6% Blue Charm (23 kills)

 

4% Green Charm (8 kills)

 

3% Crimson Charm (6 kills)

 

The most significant finding is the increase in Blue Charms.

 

I am going to get another Gargoyle task and update with more findings.

Thats hardly a significant change or indicative of anything.

 

It's just a small size size resulting in less accurate findings, the fact that from such a small sample your numbers are still all within ~3% of the main data is actually pretty on course statistically.

With a sample size thats 1/1000th of the big data test you'd expect the results to be a lot further out than ~3%

 

If anything all it says is you got lucky with a couple of extra blue charm drops than is statistically average.

Well there is certainly no correlation linked to the Mask of Granite for a specific charm based on the following numbers: 

 

Task - 284 Gargoyles

 

70.4% No Charm (200 kills)

 

12% Gold Charm (34 kills)

 

4.6% Blue Charm (13 kills)

 

5.6% Green Charm (16 kills)

 

7.4% Crimson Charm (21 kills)

 

There may not be a correlation between the mask and getting a certain type of charm, but the numbers in my first sample (be it a small one) were significantly outside 3% of a well established norm (11.6% Blue being 4.6% outside of 7%).

I said ~3% not 3% for a reason - that reason being I meant around 3% rather than exactly 3% a statement which is true as a rough average of the deviance from established percentage on your figures is in the 3% ballpark

 

The math process: You were 5% out on no charm, 2% out on gold 4% on blue 1% on green 2% on crimson aka there was about 14% deviance overall. average between 5 categories is 2.8%.

 

Therefore for the simplicity of my statement I rounded off to 3 and used the tilde symbol to indicate it was not an exact figure resulting in ~3% variance in your figures from the statistical norm and none of your variances were 'significantly' outside 3% anyway - they are all within +/-2 which is hardly a huge deviance from 3% for such a tiny sample at 198 kills for the entire sample 1% is basically 2 drops.

 

In the established norm 1% equates to 1500 data points, your entire sample being barely over 0.1% worth of their data - if as a sample it was added to the big study you'd be lucky if it made the averages shift by 0.01 of a %

With a sample size of ~200 kills, you have a margin of error that is (from a very rough calculation) around 1/sqrt(198) ~= 1/14 = 6% if my math is correct, so anything within that range shouldn't be too unexpected. That is, blues between 1% and 13% would be regarded as statistically "normal" results.

 

In fact if you average your two samples so far you get 7.46% blues, which is much closer (as you would expect, closer to the average with more samples).

 

It's always fun to test stuff though, so I hope I'm not discouraging you - only trying to provide some insight into margin of error. :)

I think, but I am not to certain, to be able to say something about significant results your smallest group needs to have at least a value of 100. In this case, you need to get at least 100 of every charm. Though I am not sure if I say this correctly, as I am basing this on psychology research, in which you want your smallest group to be at least 100 respondents (for a good research).

I think, but I am not to certain, to be able to say something about significant results your smallest group needs to have at least a value of 100. In this case, you need to get at least 100 of every charm. Though I am not sure if I say this correctly, as I am basing this on psychology research, in which you want your smallest group to be at least 100 respondents (for a good research).

 

Personally for any meaning at all of even a sort of trend to be garnered I'd say 1% needs to equal 10 or more data points aka at least 1000 data points.

For the meaning to hold real significance I'd say that'd go to 100+ minimum (aka 10,000+ data points)