Exercise: breast pump trial, 2

Question

What effect might each of these problems have on the t test?

Suggested answer

1. Treating repeated observations by the same method as independent gives us a false idea of the amount of information. This makes P values smaller than they should be.

2. Carrying out a two sample t test on paired data means that all the variation between women goes into the random variation. This tends to make the P value too large.

3. Carrying out a test when variances are unequal and distributions skew may make the P value too large.

Back to question.

Back to IAPT index.

To Martin Bland's home page.

This page maintained by Martin Bland.
Last updated: 27 February, 2009.

Back to top.