Saturday, December 4, 2010

I appreciate any help to answer the following statistical question


I appreciate any help to answer the following statistical question?
A researcher is conducting a study about the presentation of diabetes mellitus in White and Asian populations. He found that the percentages of patients presenting with weight loss were 20% and 25% in White and Asian population respectively. Which is the most appropriate statistical study to determine if there is any significant statistical difference in the study? The samples sizes are around 50s patients in each group. What I can not understand is in the T test, you will need the mean and in this question,there is no mean, it is only a percentage. Thanks for your inputs.
Mathematics - 3 Answers
Random Answers, Critics, Comments, Opinions :
1 :
two sample t-test http://ccnmtl.columbia.edu/projects/qmss/t_two.html
2 :
Generally you'd need sample sizes. You clearly didn't sample the entire white or asian populations! I think that a 2 sample t-test is probably what you'd want to use. You'd have the null hypothesis: that there's no difference between the two groups, and the alternative hypothesis: that there is You would either state a confidence level, say 95% or, calculate a p-value. But you definitely need sample sizes.
3 :
Forget t-tests! The appropriate test is a two-proportion Z-test, where: Z = [(X1/n1) -(X2/n2)]/√[phat(1-phat)(1/n1+1/n2)] Where phat = (X1 + X2)/(n1 + n2) X1 = number whites with diabetes (in this sample). X2 = number of Asians with diabetes (in this sample). Compare your calculated Z-value with the Critical Z-value found in the standard normal table with α/2. You could also do a chi-square test for independence (aka, contingency table), but that's too complicated to explain (at this time).





Read more discussions :
Posted by Diabetes Melitus at 11:12 AM