On Sun, 30 Jun 2002 01:38:37 GMT, "MPGartland" <[EMAIL PROTECTED]> wrote:
> I am familiar with the sample size determination test for continuous data, > but not discrete data such as a likert scale survey. A Likert scale gives you a sum (or average) which is continuous data, for all practical purposes. > > Basically, I am analyzing a 4 point likert scale that will be broken into 5 > cohorts. How do I determine the minimum sample size required for > significance. I will likely use an alpha of .05. Now I am a bit baffled. I think you are saying that you have one item, which is collected with Likert-headings; and you intend to use the 0-4 scores as classification. I am baffled wonder if that is really what you intend by "Likert", and because the intent -- if that is what it is -- seems odd. If you want to compare 5 groups, between all pairs of groups, you put yourself at the disadvantage of comparing the closest groups. If you have pilot data for what you intend, then it might not be too hard to make the corresponding judgements about sample. I recommend Jacob Cohen's book for ample coverage of simple power analyses, but I suggest you redo your experimental design if I have described it fully. Hope this helps. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
