We are performing a study to assess patient satisfaction with rehabilitation services. We plan to use a questionnaire with a Likert scale and the Mann-Whitney test to analyse the data.
To calculate the required sample size we plan to first calculate the sample size for a 2 sample t-test, then use Asymptotic Relative Efficiency to adjust appropriately for Mann-Whitney.
Could someone please advise us on how to go about selecting the appropriate delta in the t-test for this purpose? As the scale we are using has four discrete points, and the Mann-Whitney statistic is testing for median, does the expected difference between sample groups need to be greater than /1.0?