PDA

View Full Version : Anyone any good at statistics?


Wakefield Turntables
30-08-2012, 20:00
Can anyone answer this query i have ?

IS THERE A STATISTIC ANALYSIS WHICH CAN LOOK AT THREE VARIABLES AND PREDICT +VE/-VE LINEAR CORRELATIONS BETWEEN THE THREE VARIABLES??

I seem to think its multi-variate analysis but I need to spend some time getting my head around it? :scratch:
goraman
30-08-2012, 22:55
I could fabricate any numbers you wish and that is the art of statistics.
Wakefield Turntables
31-08-2012, 08:26
if you cant help dont post, its a waste or your time and mine.
Marco
31-08-2012, 09:37
Oi, don't be cheeky! ;)

Marco.
Martinh
31-08-2012, 10:48
Might be worth posting your query here, together with a description of what you need to acheive:

http://www.talkstats.com/forumdisplay.php/5-Statistical-Research

I'm intrigued, what is it that you are doing?

Cheers,
Wakefield Turntables
31-08-2012, 15:42
Oi, don't be cheeky! ;)

Marco.


Sorry, this is me now :spank:. Sitting is quite uncomfortable at the moment :)


Might be worth posting your query here, together with a description of what you need to acheive:

http://www.talkstats.com/forumdisplay.php/5-Statistical-Research

I'm intrigued, what is it that you are doing?

Cheers,

I assessing the effects of the scarf akin osteotomy on peak plantar pressure under the first metatarsal head. I'm doing this for my MSc, i'm in the final throws (thank the Lord!). I need to assess two X-ray angulations against a peak plantar pressure variable, I'm looking for negative or positive linear correlations between the three variables.

Hope this make sense.

ps thanks for the heads up on the website.
Barry
01-09-2012, 17:43
If you have three sets of variables to correlate, then there will be three correlation factors. I assume you have a sufficiently 'deep' set of values for each of the three variable, with which to perform a statistical analysis?
Wakefield Turntables
01-09-2012, 21:48
Barry,

define "deep"? It has various meanings. My various have very good ICC values approaching 1.0 and they are easily replicated between surgeons. I need to assess how they variable relate to each other hence linear correlation.
Barry
01-09-2012, 21:52
Barry,

define "deep"? It has various meanings. My various have very good ICC values approaching 1.0 and they are easily replicated between surgeons. I need to assess how they variable relate to each other hence linear correlation.

A large database - plenty of results. You can't do statistics on only a few results.

If you only have a few results, any coefficient of correlation derived will have enormous error of uncertainty, rendering the results meaningless.

How many values of each variable do you have?
Wakefield Turntables
02-09-2012, 09:31
I've already done a power calculation which suggest my population should be powerful enough to yield statistically significant results. I've already had one statistician look at the results, these got through the local ethics boards. My original question is for a potential error i have seen in others work, if this error turns out to be true then I can use it to discuss potential tweeks in my work for later research .