Which concept explains how outliers, missing data, or measurement errors can distort linear regression results?

Which concept explains how outliers, missing data, or measurement errors can distort linear regression results?

• A. Intercept
• B. Data Sensitivity
• C. Dummy Variables
• D. Non-Linear Terms

Answer: (B)

Outliers, missing data, and measurement errors affect how much the results of a linear regression depend on the actual data you have. This idea is data sensitivity. When data points don’t follow the general pattern, a few extreme values can pull the regression line toward themselves, changing the estimated slope and intercept even if most observations point in a different direction. Missing data reduces the information available and can bias estimates if the missingness is related to the outcome. Measurement errors introduce extra noise and can bias the relationship you’re trying to estimate, often weakening the apparent strength of the association and inflating residuals. Together, these issues show how the results can be distorted by data quality and composition. The other options don’t capture this effect. The intercept is just the baseline value when the predictor is zero and doesn’t describe sensitivity to data quality. Dummy variables are used to encode categories, not to address data quality issues. Non-linear terms model relationships that aren’t straight lines, which is a modeling choice rather than a reflection of data sensitivity.

Outliers, missing data, and measurement errors affect how much the results of a linear regression depend on the actual data you have. This idea is data sensitivity. When data points don’t follow the general pattern, a few extreme values can pull the regression line toward themselves, changing the estimated slope and intercept even if most observations point in a different direction. Missing data reduces the information available and can bias estimates if the missingness is related to the outcome. Measurement errors introduce extra noise and can bias the relationship you’re trying to estimate, often weakening the apparent strength of the association and inflating residuals. Together, these issues show how the results can be distorted by data quality and composition.

The other options don’t capture this effect. The intercept is just the baseline value when the predictor is zero and doesn’t describe sensitivity to data quality. Dummy variables are used to encode categories, not to address data quality issues. Non-linear terms model relationships that aren’t straight lines, which is a modeling choice rather than a reflection of data sensitivity.