Comparison of NPC Crew Skill Level and their Gender... Male Bias perhaps?

[Susanna]

Dont worry, we'll get back to that in a day or two, if this thread is still alive then.

Up to 75 data points now... (pitiful - I know!)

The graph has wobbled all over the place! https://www.dropbox.com/s/8mjbzbdexzr44kt/NPC Crew Profiles v2.xlsx?dl=0

I'm beginning to see what people mean by needing thousands of data points before you can iron out the RNG in this... [where is it]

 

[Kaltern]

After the last few posts from people confirming the true intention of this thread, I will give up and bow out.

All I wanted was one game, a single place, where the buzzwords of those trying to hide male sexism underneath words like 'patriarchy' and 'injustice' just didn't exist.

I guarantee this thread will eventually end up on something like Polygon or Eurogamer, and then the last vestige of real equallity will be lost forever.

I await the inevitable 'Diversity Patch', which will happen very soon. Well done.

 

[Ian Skippy]

I'm crap at math. I can tell you, logically, what should be able to be known, but I am really bad at telling you how to technically get there. So I adore it as well when it's all about cool statistics and the why's of it all.

It becomes truly interesting when you see how uncool statistics is. If you look back at my post, you'll see the p-value of the expert category is .015. That is well below the typical alpha of .05. The only reason it is not significant is due to the multiple comparison correction. In other words: if Susanna and I had only presented the expert data, it would have been significant and there would have been evidence for 'sexist bias'! I could also increase the alpha by collapsing the five separate tests into one 'is there a difference in average skill among genders' non-binomial analysis, although that analysis wouldnt have allowed for any insight into which category would be responsible for any difference. I could switch from Family Wise Error Rate correction to False Discovery Rate, and exchange an increase in Type 1 error rate for a decrease in Type 2 error rate. Fun stuff!

More cynically put: give me a dataset and a desired interpretation, and I'll write the paper that supports your claim.

Up to 75 data points now... (pitiful - I know!)

The graph has wobbled all over the place! https://www.dropbox.com/s/8mjbzbdexzr44kt/NPC Crew Profiles v2.xlsx?dl=0

I'm beginning to see what people mean by needing thousands of data points before you can iron out the RNG in this... [where is it]

Dont worry, statistical tests take sample size into account. You're just increasing your statistical power with each data point, but its not as if you really need thousands of data points to get anywhere. At some point, all you can say is that if there is a difference you failed to detect, its a difference of such a small effect size it doesnt have any practical meaning anyway.

I await the inevitable 'Diversity Patch', which will happen very soon. Well done.

Can we hold you to this prophecy?

 

[Slopey]

Closing for review.

Note that as per forum policy, discussion of politics is not permitted. A statistical anomaly based on a tiny sample of randomness, is one thing, but when it spills over into politics, that's another.