The paired-test is the same as calculating the difference between pairs and performing a 1-sample t-test against zero.
This is a question on presenting the result of the effectiveness of weight loss program. There is only one group of subject and two time point ( pre and post).
After I read some literatures, paired t test should be the most common method to show the effectiveness of the program between time points.
But, I would like to ask, is it possible/valid on using the following method?
Calculate the individual percentage change between pre and post diet, then use a one sample t test to test whether the mean individual percentage change is equal to zero.
Thx!
The paired-test is the same as calculating the difference between pairs and performing a 1-sample t-test against zero.
Last edited by Miner; 04-22-2013 at 10:13 PM.
Thx for you comment.
But, what if i am not using simple difference, but the percentage change/ relative difference? Is it still valid?
Why would you want to add the additional calculations and the added potential for error? Perform the test using differences. Then if it is significant, calculate the % change.
There is unlikely to be a statistical test that will allow you to show that the program is effective given your study design. There are any number of reasons why participants may lose weight between the two points other than the program being effective (e.g., threats to validity like history, maturation, placebo, etc...).
A paired t-test might tell you whether they lost weight, on average. But it won't tell you whether the program itself is responsible for the weight loss. To make that kind of inference you might need a better design (e.g., a randomised controlled trial, or a thoughtfully designed quasi-experiment...)
Miner (04-23-2013)
Tweet |