growthcurves.non_parametric module

growthcurves.non_parametric module#

Non-parametric fitting methods for growth curves.

This module provides non-parametric methods for growth curve analysis, including sliding window fitting and no-growth detection.

All methods operate in linear OD space (not log-transformed).

growthcurves.non_parametric.fit_non_parametric(t, N, method='sliding_window', window_points=15, smooth='fast', use_weights=False, **kwargs)[source]#

Calculate growth statistics using non-parametric methods.

This unified function supports multiple methods for calculating the maximum specific growth rate (Umax):

  • “sliding_window”: Finds maximum slope in log-transformed OD across windows

  • “spline”: Fits spline to entire curve and calculates from derivative

Parameters:

  • t (Iterable[float]) – Time array (hours)

  • N (Iterable[float]) – OD values

  • method (str, optional) – Method for calculating Umax (“sliding_window” or “spline”), by default “sliding_window”

  • window_points (int, optional) – Number of points in sliding window (for sliding_window method), by default 15

  • smooth (str, optional) –

    Smoothing mode/value for spline method, by default “fast”

    • ”fast”: notebook auto-default rule mapped to lam

    • ”slow”: GCV-selected smoothing (auto GCV)

    • float: manual lam value

  • use_weights (bool, optional) – Whether to apply OD-dependent weighting for spline method, by default False

Returns:

Dictionary with entries for - params: Model parameters (includes fit_t_min, fit_t_max, and other method-specific values) - model_type: Method used for fitting

Return type:

dict

Raises:

ValueError – Method must be “sliding_window” or “spline”. Smooth must be “fast”, “slow”, or a non-negative float.

growthcurves.non_parametric.fit_sliding_window(t, N, window_points=15, step=None, n_fits=None, **kwargs)[source]#

Calculate maximum specific growth rate using the sliding window method.

Finds the maximum specific growth rate by fitting a line to log-transformed OD N in consecutive windows using the Theil-Sen estimator, selecting the window with the steepest slope.

Parameters:

  • t – Time array (hours)

  • N – OD values (baseline-corrected, must be positive)

  • window_points – Number of points in each sliding window

  • step – Step size for sliding window (default: 1 if step is None and n_fits is None)

  • n_fits – Approximate number of fits to perform (default: None). Ignored if step is provided.

Returns:

  • slope: Slope of the linear fit in log space

    (equals specific growth rate, h⁻¹)

  • intercept: Intercept of the linear fit in log space

  • time_at_umax: Time at maximum growth rate (hours)

  • model_type: “sliding_window”

Returns None if calculation fails.

Return type:

Dict with model parameters

growthcurves.non_parametric.fit_spline(t, N, smooth='fast', use_weights=False)[source]#

Calculate maximum specific growth rate using spline fitting.

Fits a smoothing spline to log-transformed OD N and calculates the maximum specific growth rate from the spline’s derivative.

Parameters:

  • t – Time array (hours)

  • N – OD values

  • smooth – Smoothing mode/value. - “fast”: notebook auto-default rule mapped to lam - “slow”: GCV-selected smoothing (lam=None) - float: manual lam value

  • use_weights – Whether to apply OD-dependent weighting (default: False)

Returns:

  • t_knots: Spline knot points (t values)

    • spline_coeffs: Spline coefficients

    • spline_k: Spline degree (3)

    • time_at_umax: Time at maximum growth rate (hours)

    • model_type: “spline”

Returns None if calculation fails.

Return type:

Dict with model parameters