This Just In: Teenage Virgins . . . NOT! Elsewhere, Pope Suspected to Be Catholic.

Or in exactly the same words, as the case may be.

You left some key words out of your cite. “as precisely as needed”. IOW, the results will be accurate at the confidence level within the confidence interval.

So again, if you measure and find a 3 point difference but your confidence interval (at the desired confidence level) is 4 points either way, you can only say at that level of confidence that the difference among the actual population is 3±4, or the interval of (-1,7). You cannot say in such a case that you’ve shown a difference, but you certainly can’t say that you’ve shown no difference either.

The problem that was arising in my mind was that I am used to saying “No significant difference was found at alpha equal to .05. Thus the null hypothesis that no difference was present was accepted.” I understand the point you were making now.

How the fuck did a question about kids fucking turn into a pissing contest between two kids taking a class in statistics?

To nitpick, we either reject or “fail to reject” the null hypothesis (difference between two population parameters is zero).

I didn’t understand your response to my prior post. What do you mean by “Not confidence intervals, probability levels.” alpha has everything to do with confidence intervals and hypothesis testing (alpha denotes the Type I error, the probability of incorrectly rejecting the null hypothesis).

Are you confusing likelihood with confidence intervals?

I think they both are confusing sampling with hypothesis testing. It is not clear from the NY Times article which hypothesis was being tested, but let’s say it is that the pledges lower the rate of STDs. (And this would have to be in a specific time interval). It seems that this hypothesis was rejected by the study, and some p which was not given.

On the other hand, we could ask what is the rate of STDs for populations taking and not taking the pledge. Then you use sampling theory to choose the samples. We can find confidence levels around the results for both populations using well known formulas. If the confidence intervals overlap, one might be tempted to say there is no difference, but I’m not sure that this is a valid conclusion. Hypothesis testing seems a better way to proceed, especially because one might be tempted to pick and choose results for a whole bunch of values, which is not valid. so, you can’t look through the data and claim that it shows that Asian computer nerds who don’t sign the pledge have lower rates of STDs than those who do. You need a new study for that. You especially can’t discard data you don’t like, which no one in this thread has done, but which seems to be SOP for politicians.

I’m confusing everyhting with everything. I’m preparing to defend my thesis, and the nonparametric Kaplan Meier survival analysis is fucking with my head.