6.2. Workflow Pages

The GUI guides you through a 5-page workflow for curve fitting.

6.2.1. Page 1: Data Loading

Purpose: Import your data into NLSQ.

Features:

  • Load CSV files (drag-and-drop or file dialog)

  • Paste from clipboard

  • Preview data table

  • Select X and Y columns

  • View data statistics

How to use:

  1. Click “Open File” or drag a CSV onto the window

  2. Select the X column (independent variable)

  3. Select the Y column (dependent variable)

  4. Optionally select a sigma column (uncertainties)

  5. Click “Next” or Ctrl+2

Supported formats:

  • CSV (comma, tab, or semicolon separated)

  • TSV

  • Excel (xlsx) with openpyxl installed

6.2.2. Page 2: Model Selection

Purpose: Choose a mathematical model for fitting.

Features:

  • Built-in models (exponential, Gaussian, etc.)

  • Polynomial (select degree)

  • Custom Python model (advanced)

  • Model preview with formula

How to use:

  1. Select a model category (Decay, Peak, Growth, etc.)

  2. Choose a specific model

  3. View the mathematical formula

  4. Click “Next” or Ctrl+3

Built-in models:

  • Exponential decay

  • Gaussian peak

  • Lorentzian peak

  • Logistic growth

  • Power law

  • Polynomial (degree 1-10)

6.2.3. Page 3: Fitting Options

Purpose: Configure fit parameters and run the fit.

Features:

  • Workflow preset selection (Fast/Robust/Quality)

  • Initial parameter guess (auto or manual)

  • Parameter bounds (optional)

  • Advanced options (tolerances, iterations)

  • Live cost function plot

How to use:

  1. Select a preset: - Fast: Quick fit with default settings - Robust: Multi-start global optimization - Quality: Thorough search with tight tolerances

  2. Set initial guess (or use auto-estimate)

  3. Optionally set bounds on parameters

  4. Click “Run Fit” (or Ctrl+R)

  5. Watch the live cost plot during fitting

Guided vs Advanced Mode:

  • Guided: Simple preset selection

  • Advanced: Full control over tolerances, iterations, etc.

6.2.4. Page 4: Results

Purpose: View and analyze fit results.

Features:

  • Fitted parameters with uncertainties

  • Goodness-of-fit statistics (R², chi-squared)

  • Interactive fit plot

  • Residual plot

  • Correlation matrix

How to use:

  1. Review fitted parameters

  2. Examine the fit plot (zoom, pan, export)

  3. Check residuals for systematic patterns

  4. View statistics panel

  5. Click “Next” to export or “Back” to adjust

Quality indicators:

  • R² close to 1.0 indicates good fit

  • Reduced chi-squared ~1.0 with known uncertainties

  • Random residuals (no patterns)

6.2.5. Page 5: Export

Purpose: Save results in various formats.

Features:

  • ZIP session bundle (complete backup)

  • JSON (parameters and metadata)

  • CSV (data with fit curve)

  • Python code generation

  • Plot export (PNG, PDF, SVG)

Export formats:

  1. ZIP Bundle: Complete session including: - Original data - Fitted parameters - Covariance matrix - Settings and configuration - Session state for reload

  2. JSON: Structured results:

    {
  "parameters": {"A": 2.5, "k": 0.5, "c": 0.3},
  "uncertainties": {"A": 0.02, "k": 0.01, "c": 0.005},
  "statistics": {"r_squared": 0.998, "chi_squared": 45.2}
}

  3. CSV: Data table with fit values:

    x,y,y_fit,residual
0,1.0,1.02,-0.02
1,0.62,0.61,0.01
...

  4. Python Code: Reproducible script:

    from nlsq import fit
import jax.numpy as jnp


def model(x, A, k, c):
    return A * jnp.exp(-k * x) + c


popt, pcov = fit(model, xdata, ydata, p0=[2.5, 0.5, 0.3])

6.2.7. Next Steps