Which approach tends to retain extreme values?

• A. Horn's Algorithm
• B. Dixon's Range Statistic
• C. Box-plots
• D. Scatter plots

Answer: (B)

When deciding how to handle extreme observations, look for methods that are more conservative about labeling something as an outlier. Dixon's Range Statistic works by comparing the gap of the outermost value to the next value against the overall range and checking a critical value. In small samples, this criterion is fairly lenient, so an extreme value often doesn’t exceed the threshold and isn’t declared an outlier. That means the extreme observation stays in the data, influencing estimates just like any other point.

In contrast, box-plots summarize data using a 1.5 times the interquartile range rule to flag outliers, often treating those extreme points as separate from the central summary. Scatter plots simply show every data point, so extremes are always visible and retained, but as a visualization rather than a statistical outlier decision. Horn’s Algorithm and other more aggressive outlier rules tend to be stricter about excluding extremes.

So the approach that tends to keep extreme values in the dataset, rather than flagging or removing them, is Dixon's Range Statistic.