0

What data analysis is better? or fit for my data?

I've used Likert scale for my questionnaire (1-5) and I have 387 respondents.

Wilcoxon Rank Test or T-Test?

Sesese
  • 13
  • 5
  • 1
    Better by what criterion? To consider what hypothesis? How many items in your scale? This question is much too vague and broad to really answer as it stands. In addition, more focused versions of this question are almost certainly already answered -- e.g. see "Related" posts in the sidebar (-->) such as [this one](http://stats.stackexchange.com/questions/19681/when-to-use-the-wilcoxon-rank-sum-test-instead-of-the-unpaired-t-test), so even if you fixed your question enough to be answerable, it would probably be effectively a duplicate. – Glen_b Oct 02 '14 at 03:02

1 Answers1

0

A discrete variable with so few possible values can not be assumed to be distributed Gaussian, so you can't use t-test.

Also be careful using Wilcoxon tests, which one of signed-rank and rank-sum tests you have to choose depends on your data and question. If you are interested in whether median score is greater than/less than a specified value, you can use 1-sample Wilcoxon test, a slightly different approach to Wilcoxon signed-rank test. If your respondents are divided into 2 groups and you want to know whether median score of one group is greater than other one, you have to use Wilcoxon rank-sum test. If you have again 2 groups, but the respondents are somehow dependent, like they are siblings or same persons answering before and after some event, you have to use Wilcoxon signed-rank test. And if the respondents form more than 2 groups, you should consider Kruskal-Wallis test.

Macond
  • 478
  • 2
  • 9