Tutorial Menu
More About WINKS
These WINKS statistics tutorials explain the use and interpretation of standard statistical analysis techniques for Medical, Pharmaceutical, Clinical Trials, Marketing or Scientific Research. The examples include how-to instructions for WINKS SDA Version 6.0 Software. Download evaluation copy of WINKS.
Bland-Altman Plots
A Bland-Altman analysis is a way to assess agreement between two methods of clinical measurement. Using an example from a paper by Bland and Altman (1986), suppose there is a measurement (or peak expiratory flow rate/PEFR) made from using “Large” meter and a “Mini” meter. You want to know if they measure the same thing.
Two plots are used to analyze this data, a plot of identity and a Bland-Altman plot. The data follows:
SUBJECT LARGE MINI
1 494.0 512.0
2 395.0 430.0
3 516.0 520.0
4 434.0 428.0
5 476.0 500.0
6 557.0 600.0
7 413.0 364.0
8 442.0 380.0
9 650.0 658.0
10 433.0 445.0
11 417.0 432.0
12 656.0 626.0
13 267.0 260.0
14 478.0 477.0
15 178.0 259.0
16 423.0 350.0
17 427.0 451.0
The first is called a plot of identity. It is similar to a regression plot, except the line in the plot is a line based on X = Y. If the scatter of points in this plot lies near the line, it indicates that the two ways of measuring PEFR are similar. To display this plot:
Step 1: Open BAEXAMP.DBF. Select Analyze/Advanced Regression and Comparisons/Bland Altman/Line of Equality Plot.
Step 2: Select the two variables LARGE and MINI. The following plot is displayed:
From this plot of identity, you can see that the scatter of points falls close to the line.
A second method of looking at this data is with a Bland-Altman Plot. This plots the average of the two values by subject against the difference. 95% confidence limits are also calculated and displayed on the graph (Mean difference ± 2 * Standard Deviation)
To display this plot (after opening the BAEXAMP DATA):
Step 1: Select Analyze/Advanced Regression and Comparisons/Bland Altman/Bland-Altman Plot.
Step 2: Select the two variables LARGE and MINI. The following plot is displayed:
If the two measures are measuring the same thing in a unbiased way, this plot should show a random scatter of points between the upper and lower confidence limits. The plot above is indicative of a good fit. If there is a pattern to the points, there might be some biased associated with the measurements.
This procedure is available in the WINKS Professional edition.
