I have 12 results. Each one includes 157 samples. They have very different variance. I want to sort them and choose the bests based on mean. Is it accurate way to use mean function in MATLAB or something? Or I should use hypothesis test or something like that?
You should just be clear what the criterion you like to select the "best" sample. And then just identify the best sample. That's really as simple as that. You don't need a test to find the best. You would need a test only to show, for instance, that your data is sufficient to statistically distinguish the best from the second best, but that does not help you to "select the best".
Dear Mojtaba,
If you want to assess whether there is a significant difference between your results you need to use an ANOVA provided that its assumptions are met. It also depends upon whether your measurements are repeated or for different subjects.
If it is the latter, as your variances appear to be very different you should either use the Brown-Forsyth test or the Welch test for normally distributed data or the Kruskal Wallis test for non-normal data. For post hoc testing (if the first test is significant) you should either use the Games Howell test, the Dunn Bonferroni test or the Mann Whitney U test with a Bonferroni error correction.
Please see my study guides.
Dear Mojtaba
1. The criteria or statistical tests have to be developed accordingly to your research questions.
2. Determine your research questions then find suitable tests.
3. In SPSS several tests were available to differentiate between random variables/samples. Please refer to:
. Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.
.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.
Regards,
Zuhair
Dear Jochen
Thanks for your answer. I want to choose higher means. I have to say the error between variances and maximum variance of results are [ 95 ,56, 53 ,48 ,58, 8 ,77 ,32 ,33 ,32 ,45] %. Do you think use of arithmetic mean is accurate way to rank results?
The problem of ranking statistics as means is a common problem in statistics. For example, in the field of estimating (regional) indicators, it is a common and important issue.
There are only a few really convincing Ideas out there, that I'm aware of. My best guess would be to look for the term "triple goal estimators".
Article Triple-goal estimation using the U.S. Current Population Survey data
- Badges
- Science topic
- Similar topics
- Correction