Two graphs, a scatter diagram and a Normal plot.
The scatter diagram has the "Mean log CRP" on the horizontal axis numbered 1, 2, 3, 4, and 5. The vertical axis is labeled "Difference in log CRP, last minus first" and numbered -4, -2, 0, and 2. There is a horizontal line through difference = zero.
The points form a rough triangle with apex at mean = 1, difference = -0.2 and sides going from the apex to mean = 4.5, difference = +2, and to mean = 4.5, difference = -2, with points inside the triangle making it appear filled. There is one point a long way outside at mean = 2, difference = -4. As a result there is a much weaker relationship between difference and mean than in Figure 13 or Figure 14.
The Normal plot has Difference in log CRP on the vertical axis and Inverse Normal on the horizontal axis. The line of points is close to the straight line of perfect Normal distribution apart from one point at inverse Normal = -3, Difference = -4, which is noticeably below the line. Apart from this point, The line of points starts above the straight line of perfect Normal distribution, then falls below it, then rises above it again, but the departure from the straight line is slight. The line of points is closer to the straight line than it was for the untransformed or square root transformed CRP data.
Back to Clinical Biostatistics: Transformations.
This page maintained by Martin Bland.
Last updated: 8 August, 2006.