Expecon Tables Figures

Tables and Figures (expecon-tables-figures)

When to trigger

• A reader cannot see the treatment effect from your main figure without reading the text
• Tables report regression coefficients but never show the raw treatment means and their spread
• Time/round dynamics (learning, convergence, decay) are hidden inside aggregate numbers
• Error bars, units, or the inference level on a plot are ambiguous or wrong

The treatment-comparison figure is the paper's spine

At a method-defined journal, the central exhibit is almost always a comparison across treatments, and a referee should grasp the result from it alone. Design it to do three jobs.

1. Show the contrast directly. Plot the primary outcome by treatment — bars or dot-and-CI for means, but prefer showing the distribution (box/violin, or a CDF when distributions differ in shape, not just level). A CDF often reveals that a "no mean effect" masks a shift in the tails.
2. Show uncertainty at the right level. Error bars / CIs must reflect the independent unit (session/matching-group), not per-decision noise. State the level in the note. Mismatched error bars are an instant credibility hit at this journal.
3. Show dynamics when the game is repeated. A round-by-round line plot per treatment reveals learning, convergence to equilibrium, and end-game effects that an average conceals. Mark the period averaging window you use for tests.

House conventions experimentalists expect

• Raw behavior before models. Lead with a figure of actual choices (e.g., contribution distributions), then a table of treatment means with group-level SEs, then regressions for adjustment/heterogeneity. Do not open with a regression table.
• Report effect sizes, not just stars. Give the treatment difference in outcome units and as a fraction of the control mean / endowment (e.g., "+3.2 tokens, 16% of endowment"), with a CI. Significance asterisks alone are weak evidence here; many ExpEcon papers report exact p-values and CIs.
• A summary/design table. Include a treatments table (conditions × what is manipulated × n sessions/groups/subjects) so a referee can audit the design and balance at a glance.
• Self-contained notes. Each exhibit names the outcome, the unit of observation, the inference level, the test used, n at each level, and the period window. A figure a referee must decode from the body is a figure that loses trust.
• Color/encoding discipline. Keep treatment colors consistent across every figure; ensure grayscale legibility; label axes in the subjects' units (ECU) with the money conversion noted once.

Exhibit-by-exhibit plan for a typical paper

A clean ExpEcon results section usually runs a small, fixed set of exhibits — build them in this order:

Resist the urge to add a seventh table of exploratory contrasts to the main text; push it to the appendix and label it exploratory.

Execution bridge (StatsPAI / Stata MCP)

Generate exhibits from the fitted result, not by retyping numbers. Full map: execution-with-mcp. Experimental Economics is lab/field experiments; randomization inference, romano_wolf for many treatments/outcomes, and power are decisive — observational tools secondary.

• Tables: etable (multi-model) or did_summary_to_latex straight from the result_id.
• Figures: plot_from_result / enhanced_event_study_plot / event_study_table — axis units and the SE/clustering note baked in.
• Every note names the estimator + clustering and states the magnitude in interpretable units.

See a full fitted-result → exhibit chain in the JF execution walkthrough.