margot

⚠️ IMPORTANT: The margot package is undergoing refactoring as we transition to the margotverse suite of packages. The package will be split into focuse. Please expect breaking changes in upcoming releases.

MARGinal Observational Treatment-effects.¹

Causal inference requires balance across the treatments to be compared. In observational studies, such balance is not guaranteed; quantifying causality therefore requires careful, multi-step workflows.

The goal of margot is to enhance understanding of causality in observational research.

The package offers functions for:

• evaluating causal assumptions
• modelling time-series data
• reporting results
• performing sensitivity analyses

margot streamlines the estimation of (Marginal) Average Treatment Effects (ATT, ATE), but it also supports workflows for Heterogeneous Treatment Effects (CATE) (estimated via grf), as well as Longitudinal Modified Treatment Policies (estimated via lmtp). It has extensive graphical and reporting functions to ease burdens for understanding.

LMTP positivity diagnostics

For longitudinal overlap checks, margot_plot_lmtp_overlap_grid() now:

• Automatically arranges panels by shifts × waves and respects the shift order you request.
• Uses the expanded “lab” palette (grey null, blue shift_zero, distinct oranges for IPSI shifts) so panels stay visually consistent across waves.
• Applies sensible defaults for headroom and axis harmonisation—no need to hand-tune layout, ymax_harmonize, or xlim_harmonize.
• Treats the legacy layout argument as deprecated: it always reverts to the stable shifts-by-waves layout (with a CLI note if a different value is supplied).

Pair these plots with margot_interpret_lmtp_positivity() for per-wave ESS and tail diagnostics computed on uncensored weights.