Reply
Fri 11 Nov, 2011 09:19 am
I have two small groups that I want to compare scores on. One has 26 subjects, and the other 9 (this group was an ad hoc group). The groups are matched demographically. Can I use a t-test to compare the scores, or must I use the mann whitney? I am able to get the effect size from the t-test, which is why I prefer to go that route. I did run stats both ways (groups had equal variances when I ran the t-test), with the same results.
@beckadan,
What types of scores are they ? If they are physical measurements you can use t-test, but if they are psychological scales or similar you should use Mann-Whitney.
@fresco,
Thank you. The scores are psychological scales. If I use the Mann Whitney, is there an effect size calculation I can use?
@beckadan,
You need to read it up (e.g. Siegel "Non Parametric Statistics". From what I remember you put both groups together and rank all the scores. The test then measures the significance of the ranking differences with respect to group identity. The group sizes do not have to be equal.
(One reason you cannot use the t-test is that you have
ordinal data, not
interval data.)
@beckadan,
Actually, you could use a t-test if the differences between the scores are interval scales and not merely ordinal. Does the spacing between the scores have meaning across the scale?
http://www.wadsworth.com/psychology_d/templates/student_resources/workshops/stat_workshp/chose_stat/chose_stat_25.html
Stumble It!
Tweet This
Bookmark on Delicious
Share on Facebook
Share on MySpace