Bias-adjustment grid for a Cinelli & Hazlett sensitivity contour

Builds a 2-D grid over (R^2 of U with X | Z, R^2 of U with Y | X, Z) and returns the bias-adjusted estimate at each cell, ready for contour() / ggplot2::geom_contour(). The grid is dense enough (51 x 51 by default) to render smooth contours.

Usage

causal_sensitivity_grid(effect, se, df, grid_size = 51L, r2_max = 0.6)

Arguments

effect, se, df

Point estimate, SE, degrees of freedom of the causal effect.

grid_size

Integer; grid resolution per axis. Default 51L.

r2_max

Numeric in (0, 1); maximum partial-R^2 to plot. Default 0.6 (values above this are typically unrealistic for real covariates).

Value

A long-format data frame with columns r2_xu_z, r2_yu_xz, adjusted_estimate, bias, bias_factor.