If In doubt - which test is best? Para/non Para

[Gpsych]

Hi,

Which test is best to use with a small sample size? Parametric or non-parametric? (with even distribution), I've got a repeated measures design - does this reduce the limitations of the small numbers in any way? Not sure what is best to use.

Also - what should I set my p value at 0.05? when would this change to 0.001?

Thanks.

 

[ted00]

parametric has more power because it uses more information (assumptions, really), non-parametric is "safer" in terms of potential biases from assumptions not being met, as you apparently are aware. What about an exact test (depending on what the data represent)? How small of a sample size are we talking about? Is the variable continuous, binary, a rate, etc.? What are you wanting to test exactly -- location, between two groups, etc.?