Download and Install CurveExpert Basic (Windows)

How to Use CurveExpert Basic for Fast Nonlinear Regression

CurveExpert Basic is a lightweight curve-fitting tool for quickly performing nonlinear regression and visualizing results. This guide gives a concise, step-by-step workflow to fit models, evaluate results, and export outputs for reports or publications.

1. Prepare your data

1. Save data as two columns (x, y) in a CSV or tab-delimited text file.
2. Ensure consistent numeric formatting (decimal points, no thousands separators).
3. Remove or flag obvious outliers and missing-value rows.

2. Load data into CurveExpert Basic

1. Open CurveExpert Basic.
2. Use File → Open Data and select your file, or paste columns directly into the data grid.
3. Verify the correct columns are assigned to X and Y axes. Use the preview to confirm.

3. Choose candidate models

• CurveExpert Basic includes many built-in models (linear, polynomial, exponential, power, logistic, Gaussian, etc.).
• For nonlinear regression, select models that reflect the expected behavior (e.g., saturation → logistic; growth/decay → exponential).
• If unsure, pick several plausible models to compare.

4. Fit models quickly

1. Select a model from the model list.
2. Click Fit. CurveExpert will run nonlinear regression using default initial parameter guesses.
3. Review the fitted curve overlay on the scatter plot.
4. Repeat for other candidate models to compare fits.

5. Improve convergence and speed

• Use reasonable parameter initial guesses when the default fails to converge: edit parameter values in the model panel before fitting.
• Limit the number of fitted parameters where possible; simpler models fit faster and are less prone to overfitting.
• If data are noisy, consider weighting points (if available) or applying a simple smoothing prior to fitting.

6. Evaluate fit quality

• Check residuals plotted beneath the fit; look for random scatter (no systematic patterns).
• Compare statistics provided by CurveExpert Basic: R², standard error of estimate, and parameter standard errors.
• Use Akaike Information Criterion (AIC) or corrected AIC (if available) to compare non-nested models—prefer lower AIC.
• Visually inspect the overlay and residuals together with numerical metrics.

7. Handle outliers and leverage points

1. Identify points with large standardized residuals.
2. Verify whether they are measurement errors; if so, remove or correct them and refit.
3. For influential points (high leverage), test fits with and without them and report the effect.

8. Export results

• Export fitted parameter values and errors via the model results panel.
• Save plots as PNG or vector formats for publication.
• Export predicted values or residuals to CSV for further analysis.

9. Best practices and tips

• Start with simple models, then progress to more complex ones only if justified by residual patterns and improved metrics.
• Report parameter uncertainties and goodness-of-fit metrics alongside plots.
• Keep a reproducible log: original data file, model choices, initial guesses, and final parameter values.
• If multiple models explain the data similarly, prefer the simpler model (Occam’s razor) and discuss alternatives.

Quick example (workflow)

1. Load data.csv (columns: Time, Concentration).
2. Try exponential and two-phase decay models.
3. Fit exponential → poor residuals. Fit two-phase → good residuals and lower AIC.
4. Export parameters and plot for publication.

Use this workflow to get fast, reliable nonlinear regressions with CurveExpert Basic while maintaining good statistical practice and reproducibility.