What term describes the difference between the observed value of a variable and the value predicted by a statistical model?

What term describes the difference between the observed value of a variable and the value predicted by a statistical model?

• A. Deviance
• B. Residual
• C. Bias
• D. Error

Answer: (A)

The difference between what you observe and what your model predicts for each case is called the residual. It’s the observed value minus the predicted value and shows how far off the model is for that specific observation. Residuals are central for diagnosing model fit: patterns in residuals can reveal nonlinearity, heteroscedasticity, or outliers that the model isn’t capturing. Deviance, in contrast, is a summary metric used in likelihood-based models to assess overall fit by comparing the fitted model to a saturated model; it’s not the simple per-observation discrepancy. Bias refers to a systematic deviation in an estimator’s predictions on average, not the single observed–predicted gap. Error is a broader term and, in this context, can refer to the true unobserved deviation rather than the directly observed difference.

The difference between what you observe and what your model predicts for each case is called the residual. It’s the observed value minus the predicted value and shows how far off the model is for that specific observation. Residuals are central for diagnosing model fit: patterns in residuals can reveal nonlinearity, heteroscedasticity, or outliers that the model isn’t capturing.

Deviance, in contrast, is a summary metric used in likelihood-based models to assess overall fit by comparing the fitted model to a saturated model; it’s not the simple per-observation discrepancy. Bias refers to a systematic deviation in an estimator’s predictions on average, not the single observed–predicted gap. Error is a broader term and, in this context, can refer to the true unobserved deviation rather than the directly observed difference.