Basic Statistics Help

[Andy_Hollywood]

Hello

I have a problem, and seeing as I don't know a lot about stats I guess I need some help!

I have to answer the question:
With what confidence can i say that this sample came from this individual?!

Basically, I tested 11 people, they typed 'i only read' 20 times each, and i recorded latency and dwell of keystrokes (latency being length of time between successive key strokes and dwell being the amount of time a key is held down for.)

I have put all the loggged information onto the attached spreadsheet, and have calculated the mean and standard deviation. But now i don't know wher to go, so any suggestions on statistical methods regarding how i measure confidence in the sample would be great full.

The final aim of the project is to try and get a computer to tell who is at the terminal by just looking at there keystrokes.


I have been told i should be able to produce a matrix of confidence levels......but i have no idea where to start?!?

Hopefully someone can help me.

If you need more info leave a message or email me on A.Dunn@orange.net and i'll talk directly to you.

Cheers in advance for any advice

Andy

[DavidHooper]

OK, we all hate stats. But here we go:

A confidence interval (CI) is a region within which a mean (or proportion) probably falls. It is based on a sample.

That's it. Quite straightforward. For instance, I might claim that, based on this sample of 100 bananas, 3% plus-or-minus 1% of the bananas in the world are red at the 90% confidence level.

In other words, 90% of the time, a randomly chosen banana will be red with chance 3% plus-or-minus 1%. Quite a handful. Almost literally.

To set up a CI, you need three things (actually four, come to think about it). A mean (M), a standard deviation (s), a confidence level (C), the number of items (n) sampled and a set of normal distribution tables.

Find C in a set of tables - it depends on how confident you want to be.

The CI is then:
M +/- (Cs/sqrt(n))

Two caveats apply:
*Things are different for proportions so concentrate on CIs for means
*CIs only work for samples larger than about 30. If n<30 then you must use t-distribution tables which give a larger C.