Dear science community,
I added a substance to samples with 2% increment (0 (original), 2, 4, 6, 8, 10%). A regression model combined with measured variables was employed to predict the level of added substance. A RMSE of less than 1% was obtained.
Since RMSE is less 2%, is the RMSE acceptable ? Does an RMSE of less than 1% indicate unstable result in this case?
You’re right to think of RMSE as a measure of your model’s average prediction error, but "Stability" of that accuracy is a slightly different question
=>RMSE However, low RMSE alone doesn’t guarantee stability [it's more about how consistent your RMSE is across repeated samples, splits or bootstrap draws, rather than its absolute magnitude alone] -> Quantify the variability of that RMSE via Cross-validation SD (K-fold CV) or Bootstrap CIs.
=>Report both your mean RMSE and its uncertainty (SD or CI). If the upper bound remains below, say, 2%, you can confidently claim both accuracy and stability in your prediction.
- Badges
- Science topic