plot.summary.mlr: Plotting diagnostic and calibration objects resulting from...

In MatchLinReg: Combining Matching and Linear Regression for Causal Inference

Plotting diagnostic and calibration objects resulting from call to summary.mlr

Description

Diagnostic and calibration plots, inlcuding relative squared bias reduction, constrained bias estimation, bias-variance trade-off, and power analysis.

Usage

## S3 method for class 'summary.mlr'
plot(x, which = 1
  , smd.index = 1:min(10, ncol(x$smd))
  , bias.index = 1:min(10, ncol(x$bias.terms))
  , orsq.plot = c(0.01, 0.05, 0.25)
  , caption.vec = c("relative squared bias reduction", "normalized bias"
    , "standardized mean difference", "maximum bias"
    , "error components", "optimum choice", "power analysis")
  , ...)

Arguments

x	An object of class summary.mlr, typically the result of a call to summary.mlr.
which	Selection of which plots to generate: 1 = relative squared bias reduction (by term) for a single idx, 2 = bias terms vs. idx, 3 = standardized mean difference by term vs. idx, 4 = maximum bias (single-covariate, subspace, absolute) vs. idx, 5 = bias/variance/MSE plots, 6 = optimum index vs. omitted r-squared, 7 = power analysis (matched and random subsamples) vs. idx.
smd.index	Index of columns in smd.mat field of x to plot.
bias.index	Index of columns in bias.terms field of x to plot.
orsq.plot	Which values for omitted R-squared to generate plots for.
caption.vec	Character vector to be used as caption for plots. Values will be repeated if necessary if length is shorter than number of plots requested.
...	Parameters to be passed to/from other functions.

Details

Currently, 7 types of plots can be generated, as specified by the which flag: 1) relative squared bias reduction, by candidate omitted term, comparing before and after matching, 2) normalized squared bias, by candidate omitted term, vs. calibration index, 3) standardized mean difference, for all included and (candidate) omitted terms, vs. calibration index, 4) aggregate bias (single-covariate maximum, covariate-subspace maximum, and absolute maximum) vs. calibration index, 5) bias/variance/MSE vs. calibration index, at user-supplied values for omitted R-squared, 6) optimal index vs. omitted R-squared, and 7) study power vs. calibration index.

Author(s)

Alireza S. Mahani, Mansour T.A. Sharabiani

References

Link to a draft paper, documenting the supporting mathematical framework, will be provided in the next release.

MatchLinReg documentation built on Aug. 30, 2022, 5:05 p.m.