estimateEffects: Estimate Effects on the Frontier

In ChristopherLucas/MatchingFrontier: Computation of the Balance - Sample Size Frontier in Matching Methods for Causal Inference

Description

estimateEffects() is used to estimate the effect of the treatment along the entire frontier.

Usage

estimateEffects(frontier.object, my.form, prop.estimated = 1,
                mod.dependence.formula = NULL, continuous.vars = NA,
                seed = 1, model.dependence.ests = 100,
                means.as.cutpoints = TRUE, alpha=0.95)

Arguments

frontier.object	An object generated by makeFrontier().
my.form	An object of class formula (or one that can be coerced to that class). This will be passed to lm() to estimate the point estimates for the causal effect estimates across the frontier.
prop.estimated	The proportion of points on the frontier to estimate. By default, 100% of the points on the frontier are estimated. To estimate less than 100% of the points, pass the proportion to be estimated to prop.estimated (for example, .6 to estimate 60% of the points).
mod.dependence.formula	The formula used as the base formula for the Athey-Imbens model dependence estimates.
continuous.vars	All continuous control variables in mod.dependence.formula must be passed as a character vector to continuous.vars. A cutpoint for each of these variables will be estimated with segmented regression.
seed	The seed used before estimation of the effects. If prop.estimated is less than 1, this is necessary in order to replicate the exact plot.
model.dependence.ests	The number of points on the frontier to estimate. Note that for each point, a separate linear model must be fitted, so this can greater increase run time if too large. Defaults to 100 points.
means.as.cutpoints	FALSE by default. If TRUE, cutpoints are calculated as the mean instead of the breakpoint in a segmented regression. This is sometimes much faster.
alpha	0.95 by default. Confidence interval for estimated effects.

References

King, Gary, Christopher Lucas, and Richard Nielsen. "The Balance-Sample Size Frontier in Matching Methods for Causal Inference." (2015).

Examples

data(lalonde)

match.on <- colnames(lalonde)[!(colnames(lalonde) %in% c('re78',
                                'treat'))]

my.frontier <- makeFrontier(dataset = lalonde,
                            treatment = 'treat',
                            match.on = match.on)

mod.dependence.form <- as.formula(re78 ~ treat + age + black + education
                          + hispanic + married + nodegree + re74 + re75)
## Not run: 
my.estimates <- estimateEffects(my.frontier, 're78 ~ treat',
                                mod.dependence.formula = mod.dependence.form,
                                continuous.vars = c('age', 'education', 're74', 're75'),
                                prop.estimated = .1,
                                means.as.cutpoints = TRUE)

## End(Not run)

ChristopherLucas/MatchingFrontier documentation built on June 10, 2020, 6:01 p.m.