how to calculate prevalance of obesity in spss using bmi values

[arkabbot]

i am a medical student basically and we have been given this project to do , since we havent read statistics as a subject , i just want to use the simplest corelation coefficient for my project so that interpretation is easy for us..so i want to knw which corelation shud be best for me..how about the bivariate correlation , the pearson ?? will that be best in this case?

[victorxstc]

Dear arkabbot, I too think that starting with simpler tests would be better for now. However, the most basic correlation coefficients still depend on quite some factors. Without knowing their assumptions, mechanisms, and output, my (or our) answers might be quite useless to you or the future readers, not to mention that the forum guidelines needs us to help people to learn statistics, not to instruct them how to solve their stats problems without needing to learn it. Thus I suggest you study a little bit the basics of statistics before heading to running tests. Maybe starting with SPSS tutorial is good. Or you can simply google "correlation coefficients" etc.

... Our homework help policy can be found here. We mainly just want to see what you have tried so far and that you have put some effort into the problem...

[arkabbot]

ya i have been trying to learn basics , i am done with data entry and now i hav been reading correlation coefficients , i am trying to focus on pearson corelation and significance etc for now. i just needed to know if i am going in the right way..

[victorxstc]

Thats good. Why Pearson might be the right test?

[arkabbot]

coz i read that pearson is gives u corelation between variables and it looks simple to me..while multiple regression i also am trying to understand but it looks very complex..

[arkabbot]

nd multiple regression is more about predicting the effect of one variable on another while i am trying to find corelation..

[victorxstc]

Ok, so you have used these reasons for ruling out the regression, not confirming the Pearson. But there are various types of correlation coefficients, of which the Pearson is only one.

Before you can choose a test, you should know its function and its requirements (assumptions), and you must be able to recognize if the test's assumptions are met or not? If the assumptions are met and your test's function suits your design, that test might be correct. If the test does not suit your design, you should look for a whole different class of tests. If the test is good for your goal, but its assumptions are not met, you should use an alternative, usually a nonparametric alternative to that test. So in your way of learning, you should read about correlation coefficients and different types of them, and the assumptions of the different types.

So I suggest you checking the assumptions for two major types of correlation coefficients, here, (or by googling) and decide if the Pearson is good or if its nonparametric alternative, the Spearman, is better.

[arkabbot]

to check whether it is parametic or not , i tried to check skew and kurtosis to determine normal distribution
from the frequency output from SPSS...i know for normal distribution they shud be approximately 1 but these values of three variables have got me confused...is it normal distribution??
Statistics
bmi caloric intake caloric consumption
N Valid 44 44 44
Missing 0 0 0
Skewness .750 .742 -.335
Std. Error of Skewness .357 .357 .357
Kurtosis .404 -.506 -.565
Std. Error of Kurtosis .702 .702 .702