Description:
The Bland-Altman method-comparison techniques are used to compare two measurement methods of the same variable. For example, an expensive or intrusive measurement system might be compared with a one that is less expensive or less intrusive.
Suppose that a new method is developed for taking a patient’s temperature. The new method is quicker, less expensive, and less likely to cross-infect. The researchers are willing to adopt the new method if the temperature difference from the current method is no more than 0.4 degrees Fahrenheit.
To determine whether there is agreement between the two methods, the temperatures of 120 patients are taken using both methods, in immediate succession. The measurements are recorded as pairs on the spreadsheet.
In the Bland-Altman Plot and Analysis procedure, the new method is set as Method 1, and the current method for Method 2. The Calculation Options are left at the default values. On the Plot Format window, the confidence intervals of the Limits of Agreement are added to the plot.
When the procedure is run, the first report shown is the Bland-Altman plot. The difference of each pair is plotted on the Y-axis, and the average of each pair is plotted on the X-axis. The red line indicates the average difference. The points line up in rows due to the measurements being reported at only one decimal place. Thus, the differences are limited to exact one-decimal-point differences.
Before looking at the numeric reports, we can already tell that the methods will be deemed "in agreement." The outer boundaries of the plot are inside 0.4 degrees difference, which was the maximum allowed difference designated by the researchers.
The Descriptive Statistics section gives summary statistics of the two methods and the difference.
In the Bland-Altman Analysis section, we focus on the lower confidence limit of the lower limit of agreement, and the upper confidence limit of the upper limit of agreement. Because these values are inside -0.4 and 0.4, the methods are said to be "in agreement."
Although Normality of the differences is rejected in the assumption test, the assumptions plots do not show a worrisome departure from Normality.
