For EXP(1) data, 2(1.96)(D) = 3.9199 for Y and MED(n) while 2(1.96)(D) 3.51
for the 25% trimmed mean. The 25% trimmed mean may be the best of the three intervals
for these four distributions since the scaled length was small with good coverage.
The median interval was chosen so that Ln n/2 outliers are needed to drive
SE(MED(n)) to 1. (This resistant interval is not a high breakdown method since about
p
n maliciously placed outliers can drive SE(MED(n)) to zero.) Since the two resistant
intervals are easy to compute, they can be included with computer output along with
the warning to examine the data for outliers if the classical and resistant intervals differ
greatly. This application is useful since statistical consulting clients all too often obtain
their software output without plotting the data.