In addition, we investigate the effect of
extreme outliers on both separate and joint standardization techniques.
TABLE 2: NonSpurious type II error probability ([[pi].sub.[bar.D]C]) parameter for four single
extreme outlier discordancy tests.
Mild outliers were defined as viruses that had IC50 values >U and <10 times the median, and
extreme outliers as having [IC.sub.50] values >U and [greater than or equal to] 10 times the median.
All failed grades for PC and PS assays were
extreme outliers, which was similar to the AT assay and suggested accurate grading.
Nonparametric.--Mill B also had the narrowest confidence intervals from nonparametric bootstrapping (excluding the
extreme outlier) when compared with the other five mills for the complete data set derived.
These isolates respectively demonstrated a 53x and 30x reduction in zanamivir susceptibility (Table) and were
extreme outliers (data not shown).
Authors note, "to prevent outliers from influencing the analysis all variables are winsorized" and "the dependent variable was winsorized to remove the effects of
extreme outliers" and "to avoid potential problems with outliers all variables are winsorized." Similar language is also common when using trimming, dropping, or other techniques.
The outlier cutoff is taken as two times the MAD to identify moderate outliers and three times the MAD to identify
extreme outliers. This is inherently a subjective decision, but because 95.0% and 99.7% of data points are likely to be within two and three deviations from the median for a normal distribution, we thought two and three were good conservative cutoffs.
To be fair, he points out, two of the years when the Patriots played in the Super Bowl were "
extreme outliers."
They were included in the current study both because in some instances such
extreme outliers may be present in the data, and also because a goal of this research was to determine how well the various methods performed at the bounds of what might be seen in practice.
Tukey's biweight estimator has a smoothly redescending f function that prevents
extreme outliers to affect the calculation of the biweight estimates by assigning them a zero weighting.
"Aside from
extreme outliers, most often that designation and determination has more to do with our fear of reflecting upon ourselves," he says.