SAME CASES, SAME OUTCOME SCALE

Compare Regression Models

Paste a table with observed values and one or more prediction columns. The calculator reports RMSE, MSE, MAE and R² for each model on exactly the same rows, so every metric is comparable.

Actual vs predictedMultiple modelsBrowser-based

YOUR DATA

Enter observed and predicted values

Paste or upload a table with an observed column and one or more prediction columns. Use a period as the decimal mark. Up to 5 MB and 5,000 observations are accepted. Only rows with complete values for every selected column are included.

Upload a CSV or TSV file

The file is read locally in your browser.

Or paste data

One observation per row; all selected columns must be finite numbers.

Observed values

Predicted columns to compare

First row contains column namesTurn this off when the first line is data.

YOUR RESULTS

Model comparison

Metrics by model

All metrics are computed on the same complete rows and on the original outcome scale.

Model	RMSE	MSE	MAE	R²

Errors by observation

Only rows with values for every selected column are shown.

Row	Observed

How to use this calculator

Paste a table with observed values in one column and predictions from each model in additional columns.
Select the observed column and check the prediction columns you want to compare.
Click compare to see RMSE, MSE, MAE and R² side by side for the same rows.
Download the metric summary and per-row errors.

Formula and evaluation contract

Let the observed value be y and the prediction from model m be ŷ_m. For each model and the same n complete rows:

SSE_m = Σ(y − ŷ_m)²
MSE_m = SSE_m / n
RMSE_m = √MSE_m
MAE_m = Σ|y − ŷ_m| / n
R²_m = 1 − SSE_m / SST, where SST = Σ(y − ȳ)²; R² is undefined if SST = 0.

Because the metrics share the same observed values and the same rows, lower RMSE or higher R² for one model is a meaningful comparison on this evaluation set.

Interpretation and limitations

Rows with a missing value in any selected column are excluded from the comparison. The calculator shows how many rows were used.
RMSE and MAE are in the original units of the outcome. Compare them only when all models predict the same outcome on the same scale.
R² is relative to the observed mean on the evaluation rows. It can be negative and is undefined when every observed value is identical.
A better score on this evaluation set does not prove the model will perform better on new data or in production.
The current input limit is 5,000 observations and 5 MB per file.