sqrtMSE

sqrtMSE stands for the square root of the mean squared error and is commonly used in regression analysis as a measure of predictive accuracy. It is defined as sqrtMSE = sqrt( (1/n) ∑_{i=1}^n (y_i − ŷ_i)^2 ), where y_i are observed values and ŷ_i are predicted values. The result shares the same units as the target variable, which aids interpretation in practical terms.

Because sqrtMSE equals the root mean squared error (RMSE), it inherits its properties: it is non-negative, with zero indicating perfect predictions on the data (in the given sample). It is sensitive to outliers due to the squaring of errors and thus emphasizes large deviations more than absolute-error metrics like MAE (mean absolute error).

sqrtMSE is widely used to compare regression models, tune hyperparameters, and report model performance on held-out data or across cross-validation folds. It can be computed from any set of predictions and actuals, and is meaningful only when the same data partition and target scale are used.

Compared with other error metrics, sqrtMSE provides direct interpretability in the response variable’s units. However, it does not reveal bias direction or error distribution; residual plots or additional metrics may be used to supplement its interpretation. Related metrics include MSE, RMSE, MAE, and MAPE.