Hello everyone,
I have scores on different scales of a sample of children, and I want to compare the mean scores on the different scales of this sample. For instance, my hypothesis is that children have higher mean scores in Vitality than in Tension. Can I use a paired t-test?
Do I have to preliminary test the normality of the distribution? How?
Thanks to everyone, in advance.
If the distributions are approximately normal, then you can use a paired t-test. You can just eyeball the distributions and the descriptive statistics. (I consider the statistical tests for normality too stringent for most parametric tests.) If the distributions do not look normal, then you can transform the data; which technique you use depends on the distribution. Get back to us after you have looked at the distributions & feel you need more guidance.
However, this only works in case that the possible ranges of the scores to compare are the same.
As I understood from your hypothesis, you should use an independent sample t-test instead paired sample test. It will tell you to see whether vitality is higher or tension?
Naeem Aslam A paired t-test is more appropriate because the two measures being compared are generated by the same respondents.
Rhianon Allen, I think that if the measure is the same and administered at two-time points then a paired t-test would be helpful. Here the measures are different and how different measures can be compared across time.
I may suggest, that total scores of the two measures are compared in a population by using an independent sample t-test. This is my understanding
- Badges
- Science topic
- Similar topics
- Medicine
- Epi
- Age Groups
- Children