Skip to content

FILE: docs/api/scipy.md

SciPy API Reference

Functions for statistical testing, effect sizes, multiple comparisons correction, and curve fitting.

compare_groups(a, b, alpha=0.05, paired=False, verbose=True)

Compare two groups using the appropriate statistical test. Automatically checks normality and selects t‑test or Mann‑Whitney/Wilcoxon.

Parameters: - a, b – array‑like, the two groups to compare - alpha – significance level (default 0.05) - paired – whether the groups are paired observations - verbose – print the full report

Returns: Dict with keys: - test_name – name of the test used - statistic – test statistic - pvalue – p‑value - significant – boolean (p < alpha) - alpha – significance level - effect_size – Cohen's d - effect_label – 'negligible', 'small', 'medium', 'large' - normality_a, normality_b – bool - n_a, n_b – sample sizes - interpretation – plain‑English summary

Returns None if inputs are invalid.

Example: 

result = dk.compare_groups(group1, group2)
print(result['interpretation'])

result = dk.compare_groups(before, after, paired=True)

cohen_d(a, b)

Compute Cohen's d effect size for two independent groups.

Parameters: - a, b – array‑like

Returns: (d_value, label) where label is 'negligible', 'small', 'medium', or 'large'. Returns (None, None) if pooled standard deviation is zero.

Example: 

d, label = dk.cohen_d(treatment, control)

correct_pvalues(pvalues, method='fdr_bh', alpha=0.05, verbose=True)

Apply multiple comparisons correction to a list of p‑values.

Parameters: - pvalues – list or array of p‑values - method – 'bonferroni', 'fdr_bh' (Benjamini‑Hochberg), 'fdr_by', 'holm' - alpha – family‑wise error rate - verbose – print a table of original vs corrected p‑values

Returns: Dict with keys: - reject – boolean array - pvalues_corrected – array - method, alpha - n_rejected, n_total

Returns None if statsmodels is not installed.

Example: 

corrected = dk.correct_pvalues(pvals, method='bonferroni')
print(f"Rejected {corrected['n_rejected']} hypotheses")

fit_and_plot(model_fn, x, y, p0=None, plot_residuals=True, figsize=(12, 4), show=True, save=None, dpi=150, verbose=True)

Fit a custom model function to data and plot the fit (and optionally residuals).

Parameters: - model_fn – function f(x, *params) → y - x, y – array‑like, data points - p0 – initial guess for parameters (optional) - plot_residuals – show residuals plot (right panel) - figsize – figure size (width, height) - show – call plt.show() - save – file path to save figure - dpi – resolution - verbose – print parameter estimates and standard errors

Returns: Dict with keys: - popt – optimal parameters - pcov – covariance matrix - perr – standard errors - residuals – array - r_squared – R² value - fig – matplotlib Figure - axes – list of Axes objects

Returns None if fitting fails.

Example: 

def linear(x, a, b):
    return a * x + b

result = dk.fit_and_plot(linear, x, y)
print(f"Slope: {result['popt'][0]:.3f}, R²: {result['r_squared']:.3f}")