Which term refers to extreme values that distort results?

Which term refers to extreme values that distort results?

• A. Outliers
• B. One-Hot Encoding
• C. Informative Missingness
• D. Longitudinal data

Answer: (A)

Extreme values that lie far from the rest distort results. These observations, known as outliers, can pull the mean toward them, inflate variance, and skew relationships in ways that misrepresent the typical pattern in the data. Because many statistical methods assume data cluster around a central value and that relationships are roughly linear, a single outlier can shift estimates, alter regression slopes, and change p-values or confidence intervals. Detecting them with simple rules like the IQR method or standardized scores helps decide whether they are data entry errors, rare but real observations, or legitimate signals to model differently. If they’re errors, correct or remove them; if they’re real but influential, use robust methods or transformations that lessen their impact, or compare results with and without them to see how conclusions change. The other terms refer to different ideas: converting categories into binary indicators, missing data patterns that carry information, and data collected over time for the same subjects.

Extreme values that lie far from the rest distort results. These observations, known as outliers, can pull the mean toward them, inflate variance, and skew relationships in ways that misrepresent the typical pattern in the data. Because many statistical methods assume data cluster around a central value and that relationships are roughly linear, a single outlier can shift estimates, alter regression slopes, and change p-values or confidence intervals. Detecting them with simple rules like the IQR method or standardized scores helps decide whether they are data entry errors, rare but real observations, or legitimate signals to model differently. If they’re errors, correct or remove them; if they’re real but influential, use robust methods or transformations that lessen their impact, or compare results with and without them to see how conclusions change. The other terms refer to different ideas: converting categories into binary indicators, missing data patterns that carry information, and data collected over time for the same subjects.