pylimma.fit_f_dist_unequal_df1

pylimma.fit_f_dist_unequal_df1(x, df1, covariate=None, span=None, robust=False, prior_weights=None)[source]

Fit scaled F-distribution with unequal df1 values.

Robust estimation of the parameters of a scaled F-distribution when df1 varies substantially between observations. Uses maximum likelihood estimation with inverse-trigamma weighting.

This version is preferred when df1 values are small or vary across genes, such as from edgeR quasi-likelihood pipelines.

Parameters:

  • x (array_like) – Sample variances. Should be positive.

  • df1 (array_like) – Numerator degrees of freedom for each variance (can vary per gene).

  • covariate (array_like, optional) – If provided, allows the scale to vary as a function of the covariate.

  • span (float, optional) – LOWESS span for covariate trend. If None, chosen automatically.

  • robust (bool, default False) – Use FDR-based outlier detection and re-weighting.

  • prior_weights (array_like, optional) – Prior weights for each observation.

Returns:

scalefloat or ndarray

Estimated prior variance (s0^2).

df2float

Estimated prior degrees of freedom (d0).

df2_shrunkndarray, optional

Gene-wise shrunken df2 (only if robust=True and outliers found).

Return type:

dict

Notes

Key differences from fit_f_dist: - Uses inverse-trigamma weighting: w = 1/trigamma(df1/2) - Uses maximum likelihood optimization instead of method of moments - Better handles small or varying df1 values

References

Smyth, G. K. and Chen, L. (2024). limma package source code.