Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • New on SSC: rcsplot — publication-ready restricted cubic spline plots with one command

    I am pleased to announce that rcsplot is now available from the SSC archive. To install:

    ssc install rcsplot
    rcsplot fits a restricted cubic spline (RCS) model for a continuous exposure and produces a publication-ready dose–response curve in a single command. It works with logistic, logit, poisson, stcox, and regress, and returns odds ratios, hazard ratios, rate ratios, or beta coefficients depending on the model.

    Key features:
    • Reference-centred curves — the curve is anchored at a user-specified reference (median by default), so the effect estimate is exactly 1 (or 0 for linear models) at that point.
    • Two P-values reported — overall association and non-linearity, printed in the output and optionally on the plot.
    • Built-in trimming — observations outside the 1st–99th percentile of the exposure are dropped before fitting, and the plotted x-range matches the modelled sample by default.
    • Survey support — works with svy, including domain estimation via svyopts(, subpop(if ...)), with knots and reference computed on the trimmed subpopulation.
    • Publication-ready graphics — optional histogram overlay, shaded confidence bands in Nature-journal style (ciband), and adjustable band opacity (cibopacity()).
    A typical call:

    webuse nhanes2, clear

    rcsplot logistic highbp zinc age i.sex bmi, knots(4) ciband histogram

    This fits a 4-knot RCS logistic model for zinc (adjusted for age, sex, and BMI), plots the OR curve with a shaded 95% CI, overlays a histogram of zinc.

    The command is documented with a full help file (help rcsplot).

    Feedback and bug reports are welcome.
    Last edited by Zumin Shi; 07 Oct 2026, 13:25.
Working...
X