ToukaOone posted...
What do you mean by the false positive/negative rate for the Kinsey scale? I'm not familiar with all stat lingo. Is that how many people completed the question?
No uhhh.
Let's say that we find a bunch of people who we don't know are hetero/homosexual, and we find that consistently homosexuals score high on the Kinsey scale and heterosexuals score low.
How many heterosexuals score high (false positive) and how many homosexuals score low (false negative) ...?
The way to think about this is that there are TRUE positives (Testing for homosexuality -> Are homosexual), TRUE negatives (Testing for not homosexual -> not homosexual) and false versions of the above (test homosexual -> not homosexual, test for not homosexual -> not not homosexual aka homosexual). We need to know how well it correlates as well as how well it doesn't correlate to know how good a test that is.
I was able to use pretty much every survey that I received since the majority of the surveys' questions were isolated from one another. For instance, if I had more people tell me how masculine or feminine "playing basketball" was then "wearing pink clothing," I can still individual correlate those behaviors with sexual orientation. They don't rely on one another.
But wouldn't we exactly expect to find one or two non correlated results, since the condition is that there's a one in twenty chance of meeting the p value and still being the null hypothesis? Or am I misunderstanding?
Okay, well, we'd have to get into the accuracy of self-reporting sexual orientation, then. It's pretty hard to quantify that. But, in general, I would say that due to the privacy of the survey and the nature of the question ("What sex do you sexually desire?"), I have to just rely on what they are telling me.
But, I would also say that you can't actually test for homosexuality in a way other than analyzing sexual desire. Like I've said in the past, you can't look at behavior. I can engage in oral sex with a man for many reasons that don't affect my sexual orientation of being "straight." As an example, I could work in gay-for-pay, where men who perform in gay scenes receive almost ten times as much money as men who perform in straight scenes.
There really isn't a more effective way of looking at sexual orientation than asking people who they desire. If we try to say that there is a more effective way--by looking at behavior or past experiences--than we're literally saying that behavior is indicative of orientation. Which is pretty much what this paper argued against.
As for your final point, I could have judged statistically significance at a 99% confidence interval (p < .01), or even 99.9% confidence (p < .001). In fact, most of these significance values were less than .01, if I recall correctly. With the exception of maybe 1 or 2. But it's not a 1 in 20 chance of meeting the p-value verbatim, I don't believe. It's still based on standard deviations. I can say with 95% confidence these figures are correct (if this was a perfect random sample, which it wasn't, of course). That's what that figure means.
Also, the null hypothesis was that there was no correlation.
--
Women come and go, but Metal Gear is forever.
Perfect Odds A Cappella ~ http://i43.tinypic.com/34pdhc7.jpg