pySEQTarget 0.13.9

• Adds SEs and modifiable labels to the risk output.

pySEQTarget 0.13.8

Small fixes, backend robustness, and bootstrap speedups.

pySEQTarget 0.13.7

• Bump a worflow in CI
• Warn when numerator and denominator weight models use identical covariates
• Adds real tests for several options (like just implemented for SEQTaRget)
• Adds an additional attempt to fix the readthedocs navbar height
• Adds warm starts to the glum bootstrap fits
• Cache the patsy design_info across bootstrap outcome fits
• Use integer IDs through the bootstrap path
• Skip the pl.from_pandas round-trip in the weighted fit block
• Use a fixed default seed when none is supplied
• Add JAX backend for GLMs

pySEQTarget 0.13.6

This adds glum and scikit-survival backends to speed up the GLM and Cox fits. The default fitting packages are left as statsmodels and lifelines.

I've also added as similar output tables to SEQTaRget as I could manage.

pySEQTarget 0.13.5

This adds the two most recent additions to the R package:

• Add risk_times option

• Ensure stable factor encoding for categorical time-varying covariates in the outcome model

And

• Reduce the height and padding of the piccolo_theme top navigation bar in the readthedocs site

pySEQTarget 0.13.4

• Add warm starts for bootstrap GLM fits
• Print dataset summary information during the run when new SEQopts() verbose=True

pySEQTarget 0.13.3

• Pre-compute spline knots from full dataset to ensure consistency across bootstrap replications
• Replace squared spline term with the spline basis itself
• Add followup_spline_df parameter to control spline degrees of freedom
• Add Modelling Followup with a Natural Cubic Spline section to docs/vignettes/more_advanced_models.md

pySEQTarget 0.13.2

• Add expand_only option to SEQopts to return expanded dataset without running full analysis.

pySEQTarget 0.13.1

• Add Tom to maintainers

pySEQTarget 0.13.0

• Preserve eligible_col during expansion when referenced by weight_eligible_colnames
• Validate weight_eligible_colnames entries exist in data upfront
• Skip bootstrap replicates that fail with singular Hessian
• Silence non-interactive backend plt.show warnings in test suite
• Bump astral-sh/setup-uv to v8.1.0