Dear Christopher D. McDermott ,

Yes, this does sound like an experiment. If you try using a dimension reduction technique on your Likert response questions you might be able to use a more sensitive and sophisticated design, such as a repeated measures ANOVA. Please see my study guides on scale reliability analysis.

Thank you Peter.

Do you think a Mann-Whitney U test would be the most appropriate considering I will use it to compare the differences between two independent groups when my dependent variable is ordinal (Likert)

Having spoken to a few other academics, they have suggested that since the whole experiment will be undertaken all at once, the use of a control group will not be needed. The reason given was that I could simply compare the results of the intervention groups pre and post test answers, to ascertain if any significant difference has occured, as a result of using the software.

Would anyone like to comment on this, and advise whether this is correct, and I could just use the same methodology I employed in the pilot study.

flow diagram attached

thanks

Dear Sir , Christopher Mcdermott

Check the normality of data, if it is normaally distributed, then used pared t-test if not used Mann-Whitney U test

Hari Prasad Upadhyay Thank you for your response and article. The article was very useful and helped me better understand how to choose a statistical method. Could you provide comment on my message above, asking if the use of a control group in my experiment is needed.

many thanks

As my knowledge it is not that of experiment,

for the case control study case (people with disease) or people who get intervention, and control should be free (with out intervention). if you use totally random process of selection of case and control then only your design is experimental if not it is Quasi experimental design.

Mr. Christopher Mcdermott can i got your email or email to me : [email protected]

I have a question for Christopher D. McDermott:

I am doing a similar study to your pilot study and I would really appreciate it if you could tell me why you used a Wilcoxon signed-rank test.

I also have likert questions and I was about to use a paired samples t test but then I realised that I have ordinal data and Im still trying to find out which kind of test suffices. Is wilcoxon ok?

Many thanks :)

Georgios Mitsostergios I'm sure there are more qualified people on here who may be able to help. In my case, my data was collected using Likert scales and was, therefore, not normally distributed. Since I was testing the same group Pre/Post, and the assumption for t-test was violated, I used wilcoxon signed-rank to test for significance in median differences.