In regression, when are residuals of two observations considered to be autocorrelated?

• A. Autocorrelation
• B. Bar chart
• C. ANOVA
• D. β-level

Answer: (A)

Residual autocorrelation occurs when the error terms from different observations are not independent. In regression, we expect residuals to be independent; autocorrelation means e_i and e_j are correlated for i ≠ j, which is common with sequential or time-ordered data. When this happens, the usual standard errors and test statistics can be biased, making significance tests unreliable and confidence intervals too narrow or too wide. A common way to diagnose first-order autocorrelation is the Durbin-Watson statistic, where values near 2 suggest little or no autocorrelation, and values far from 2 indicate positive or negative autocorrelation. Other terms like a bar chart, ANOVA, or β-level refer to different concepts and don’t describe residual correlation.