0

I've conducted a survey for my thesis where I ask a series of questions of the format:

A statement relating to the experiment

Then the participants answer strongly disagree (1) --> strongly agree (5)

I have 14 responses and two groups, one control and one experimental (7) in each. With the answers to each of these questions, I have checked for statistical significance in each question by use of a Z-test, however I was reading this morning that my data may be ordinal and small sample size, in which case I should use a T-test.

Is my initial approach valid still?

Owen Morgan
  • 101
  • 1
  • See https://stats.stackexchange.com/questions/85804/choosing-between-z-test-and-t-test or https://stats.stackexchange.com/questions/61284/t-tests-vs-z-tests , but the more general question is if you should use *any* of those tests with this data. – Tim Aug 24 '17 at 11:19
  • Wouldn't fisher's exact test be a better fit here, since your answers are categorical and sample size is small? – TenaliRaman Aug 24 '17 at 11:25
  • I don't understand "one control and one experimental in each". Are you saying that controls and treatments are paired? Or did you randomly assign 14 people to two groups? – Placidia Aug 24 '17 at 11:34
  • random assignment of people to two groups, so participants either did the experiment or the control. – Owen Morgan Aug 24 '17 at 11:36
  • Why would *ordinal data* suggest a t-test? – Glen_b Aug 24 '17 at 12:29
  • Echoing Glen_b 's response: Your dependent variables are ordinal in nature, so you should use tests appropriate for ordinal responses. Some you might look into: Cochran-Armitage test (perfect for your case); Mann-Whitney test (common and simple, but some people don't think it should it be used for discrete data with many ties, like your Likert response data); ordinal regression (in some software packages, not as scary as it sounds, but might be overkill for what you need). In any case, t-test or z-test are probably not appropriate. – Sal Mangiafico Aug 24 '17 at 13:17
  • HOW many respondents and how many question items do you have ? What are your question items and what are trying to estimate ? It seems your scoring is unjustified. –  Aug 24 '17 at 14:33

0 Answers0