Question 2: How can we interpret the antilog of the difference and its confidence interval?
The antilog of the difference between the means on the log scale is the ratio of the means on the natural scale. The antilog of a mean is the geometric mean, so the antilog of the difference is the ratio of two geometric means. So the point estimate for the ratio of the geometric mean albumin for COAD patients is twice that for controls, with interval estimate 1.4 to 3.0.
Back to Exercise: Transformations.
This page maintained by Martin Bland.
Last updated: 27 July, 2009.