summary.lmw_est_aipw {lmw}

Extract effect estimates and standard errors from lmw_est fits

Description

summary() computes the treatment effect and potential outcome mean estimates from the supplied lmw_est object. It functions similarly to summary.lm() in producing estimate tables with the estimates, standard errors, t-statistics, and p-values. Other model statistics can be additionally requested.

Usage

## S3 method for class 'lmw_est_aipw'
summary(object, model = FALSE, ci = TRUE, alpha = 0.05, ...)

## S3 method for class 'lmw_est'
summary(object, model = FALSE, ci = TRUE, alpha = 0.05, ...)

Arguments

object

an lmw_est object; the output of a call to lmw_est.

model

logical; whether to produce a coefficient table for the outcome model coefficients. Note that these values should not be interpreted or reported so they are not produced by default.

ci

logical; whether to include confidence intervals in the output.

alpha

when ci = TRUE, the alpha value used to compute the critical test statistic for the confidence interval; equivalently, 1 minus the confidence level (e.g., for a 99% confidence interval, alpha = .01 should be specified). Default is .05 for a 95% confidence interval.

...

ignored.

Details

summary.lmw_est() produces a table of treatment effect estimates corresponding to all possible pairwise contrasts between the treatment levels. These treatment effects generalize to the population implied by the regression weights, which depends on the supplied estimand, whether sampling weights were provided, and which of the MRI or URI models was requested. The treatment effects are computed using linear contrasts of the outcome model coefficients.

When method = "MRI", the potential outcome mean estimates are also reported. These correspond to the potential outcome means in the population implied by the regression weights. When method = "URI", only the treatment effects are estimated; the model-implied outcome means do not correspond to the potential outcome means for the population implied by the regression weights. That is, while the treatment effect generalizes to the population defined by the regression weights, the estimated potential outcome means do not and so are not reported.

When model = TRUE, the model coefficients and their tests statistics are additionally produced. It is inappropriate to interpret or report these values as they have no causal interpretation. This is especially true when using AIPW, as the model coefficients do not incorporate the augmentation terms.

Value

A summary.lmw_est object with the following components:

call

the original call to lmw_est()

means

a matrix containing the estimated potential outcome means, their standard errors, confidence interval limits (if requested with ci = TRUE), t-statistics, and p-values. Omitted when method = "URI" or fixef is not NULL and for lmw_iv objects.

coefficients

a matrix containing the treatment effect estimates and their standard errors, t-statistics, and p-values.When ci = TRUE, the confidence limits ⁠"95%" CI L⁠ (lower) and ⁠"95%" CI U⁠ (upper) will be included between the standard error and t-statistic columns. When AIPW is used, z-statistics and z-tests are reported instead.

model.coefficients

when model = TRUE, the coefficient table of the model coefficients, which has the same columns as coefficients.

aliased

when model = TRUE, a named logical vector showing if the original coefficients are aliased (i.e., NA).

sigma, df, r.squared, adj.r.squared

the residual standard deviation, degrees of freedom components, R-squared, and adjusted R-squared. See summary.lm(). When AIPW is used, sigma and df are omitted.

Other components containing information for printing are also included.

See Also

lmw_est() for fitting the outcome regression model, summary.lm() for more information on the output components

Examples

# See examples at `help("lmw_est")`