(no subject)

posted by [identity profile] theinquisitor.livejournal.com at 06:42am on 03/11/2003
Sad to say (actually, not very), I probably count as one now. Certainly if job application forms are to be believed.
[parent entry] [link] [reply]
emperor: (Default)

(no subject)

posted by [personal profile] emperor at 07:00am on 03/11/2003
Well in that case, can you help me with this Simple Statistics Question?

First, the simplified form: I have a series of observations. How do I assess whether their mean is significant?

The actualy problem: I have 3 rhinos, which have each been anaesthetised a fair few times. In each case, the time to onset of anaesthesia has been recorded in a proportion of cases. How do I tell whether the values I have tell me anything useful? (i.e. mean time of onset might be 5 minutes, but is this significant?)
[link] [parent] [reply]
 

(no subject)

posted by [identity profile] theinquisitor.livejournal.com at 08:04am on 03/11/2003
Hmmm... What you want, (assuming that your distribution is anything resembling normal) is a t-test.

If I remember correctly (and I never did these at A Level, for no reason I'm aware of) you normalise to a variance of 1 (obtaining global variance estimates from sample variance is a case of multiplying by n/(n-1), IIRC), and a mean of 0, and then pull out a lookup table, which will give the boundary values beyond which that mean is significant at the (say) 5% level.

I'm assuming here that you're adopting the standard approach to hypothesis testing, i.e. you have a control sample, and thus a 'typical' mean, and you are trying to verify the assertion that a particular sample is (say() greater than the mean, and that there is less than a 5% chance this could have come about randomly.

Hmmm... That's almost certainly useless to you, but I think the critical point is to look up a t-test in a stats textbook of whatever level you have the background for.

IIRC, you're trying to verify the assertion that the onset time increases with repeated applications. Were I attempting to do this without doing some serious reading beforehand, I'd compare likelihoods of a fixed mean, and one which decreased with time. Still using a t-test, but 'normalise to mean 0' takes on a slightly different meaning.

It's really quite hard to explain maths by email, so I can't imagine this makes much sense. Still, hope it helps a little.
[link] [parent] [reply]
 

(no subject)

posted by [identity profile] teleute.livejournal.com at 07:28pm on 03/11/2003
I've just discovered two interesting things. 1. t-tests are what the American's call z-scores (which is why I've been wondering where the t-tests are, and what the hell z-scores are, having never done stats at school). 2. and more importantly for Matthew, I can do this too, so *waves at matthew* if you need an extra helping hand, I can not only do this, but I've taught it to people before!
[link] [parent] [reply]

October

SunMonTueWedThuFriSat
      1
 
 2
 
 3
 
 4
 
5
 
 6
 
 7
 
 8
 
 9
 
 10
 
 11
 
12
 
 13
 
 14
 
 15
 
 16
 
 17
 
 18
 
19
 
 20
 
 21
 
 22
 
 23
 
 24
 
 25
1
26
 
 27
 
 28
 
 29
 
 30
 
 31