ps.cont: Gradient boosted propensity score estimation for continuous...

In twangContinuous: Toolkit for Weighting and Analysis of Nonequivalent Groups - Continuous Exposures

Description

'ps.cont' calculates propensity scores using gradient boosted regression and provides diagnostics of the resulting propensity scores.

Usage

ps.cont(
  formula,
  data,
  n.trees = 10000,
  interaction.depth = 3,
  shrinkage = 0.01,
  bag.fraction = 1,
  sampw = NULL,
  print.level = 2,
  verbose = FALSE,
  stop.method = "wcor",
  treat.as.cont = FALSE,
  ...
)

Arguments

formula	An object of class [formula]: a symbolic description of the propensity score model to be fit with the treatment variable on the left side of the formula and the potential confounding variables on the right side.
data	A dataset that includes the treatment as well as the potential confounding variables.
n.trees	Number of gbm iterations passed on to [gbm]. Default: 10000.
interaction.depth	A positive integer denoting the tree depth used in gradient boosting. Default: 3.
shrinkage	A numeric value between 0 and 1 denoting the learning rate. See [gbm] for more details. Default: 0.01.
bag.fraction	A numeric value between 0 and 1 denoting the fraction of the observations randomly selected in each iteration of the gradient boosting algorithm to propose the next tree. See [gbm] for more details. Default: 1.0.
sampw	Optional sampling weights.
print.level	The amount of detail to print to the screen. Default: 2.
verbose	If 'TRUE', lots of information will be printed to monitor the the progress of the fitting. Default: 'FALSE'.
stop.method	A method or methods of measuring and summarizing balance across pretreatment variables. Current options are 'wcor', the weighted Pearson correlation, summarized by using the mean across the pretreatment variables. Default: 'wcor'.
treat.as.cont	Used as a check on whether the exposure has greater than five levels. If it does not and treat.as.cont=FALSE, an error will be produced. Default: FALSE
...	Additional arguments that are passed to ps function.

Value

Returns an object of class 'ps.cont', a list containing

* 'gbm.obj' The returned [gbm] object.

* 'treat' The treatment variable.

* 'desc' A list containing balance tables for each method selected in 'stop.methods'. Includes a component for the unweighted analysis names “unw”. Each 'desc' component includes a list with the following components

- 'ess' The effective sample size.

- 'n' The number of subjects.

- 'max.wcor' The largest weighted correlation across the covariates.

- 'mean.wcor' The average weighted correlation across the covariates.

- 'rms.wcor' The root mean square of the absolute weighted correlations across the covariates.

- 'bal.tab' a (potentially large) table summarizing the quality of the weights for balancing the distribution of the pretreatment covariates. This table is best extracted using the [bal.table] method. See the help for [bal.table] for details.

- 'n.trees' The estimated optimal number of [gbm] iterations to optimize the loss function.

* 'ps.den' Denominator values for the propensity score weights.

* 'ps.num' Numerator values for the propensity score weights.

* 'w' The propensity score weights. If sampling weights are given then these are incorporated into these weights.

* 'datestamp' Records the date of the analysis.

* 'parameters' Saves the 'ps.cont' call.

* 'alerts' Text containing any warnings accumulated during the estimation.

* 'iters' A sequence of iterations used in the GBM fits used by 'plot' function.

* 'balance' The balance measures for the pretreatment covariates used in plotting.

* 'sampw' The sampling weights as specified in the 'sampw' argument.

* 'preds' Predicted values based on the propensity score model.

* 'covariates' Data frame containing the covariates used in the propensity score model.

* 'n.trees' Maximum number of trees considered in GBM fit.

* 'data' Data as specified in the 'data' argument.

References

Zhu, Y., Coffman, D. L., & Ghosh, D. (2015). A boosting algorithm for estimating generalized propensity scores with continuous treatments. *Journal of Causal Inference*, 3(1), 25-40. doi: 10.1515/jci-2014-0022

See Also

[gbm], [plot.ps.cont], [bal.table], [summary.ps.cont]

Examples

  ## Not run: test.mod <- ps.cont(tss_0 ~ sfs8p_0 + sati_0 + sp_sm_0
          + recov_0 + subsgrps_n + treat, data=dat
## End(Not run)

twangContinuous documentation built on Feb. 26, 2021, 5:09 p.m.