WeightIt (version 0.8.0)

method_twang: Propensity Score Weighting Using Generalized Boosted Models with TWANG

Description

This page explains the details of estimating weights from generalized boosted model-based propensity scores by setting method = "twang" in the call to weightit or weightitMSM. This method can be used with binary, multinomial, and continuous treatments.

In general, this method relies on estimating propensity scores using generalized boosted modeling and then converting those propensity scores into weights using a formula that depends on the desired estimand. The algorithm involves choosing a balance criterion to optimize so that balance, rather than prediction, is prioritized. For binary and multinomial treatments, this method relies on ps from the twang package. For continuous treatments, this method relies on ps.cont from WeightIt.

See the Note section at method_gbm for a discussion of how this method differs from method = "gbm". In general, using method = "gbm" is preferred because it is faster and more flexible. method = "twang" will likely become defunct in future versions.

Binary Treatments

For binary treatments, this method estimates the propensity scores using ps. The following estimands are allowed: ATE, ATT, and ATC. The weights for the ATE, ATT, and ATC are computed from the estimated propensity scores using the standard formulas. When the estimand is the ATE, the return propensity score is the probability of being in the "second" treatment group, i.e., levels(factor(treat))[2]; when the estimand is the ATC, the returned propensity score is the probability of being in the control (i.e., non-focal) group.

Multinomial Treatments

For multinomial treatments, this method estimates the propensity scores using multiple calls to ps (which is the same thing that mnps in twang does behind the scenes). The following estimands are allowed: ATE and ATT. The weights are computed from the estimated propensity scores using the standard formulas.

Continuous Treatments

For continuous treatments, the generalized propensity score is estimated using ps.cont.

Longitudinal Treatments

For longitudinal treatments, the weights are the product of the weights estimated at each time point, which is what iptw in twang does behind the scenes.

Sampling Weights

Sampling weights are supported through s.weights in all scenarios.

Missing Data

In the presence of missing data, the following value(s) for missing are allowed:

"ind" (default)

First, for each variable with missingness, a new missingness indicator variable is created which takes the value 1 if the original covariate is NA and 0 otherwise. The missingness indicators are added to the model formula as main effects. The weight estimation then proceeds with this new formula and set of covariates using surrogate splitting as described below. The covariates output in the resulting weightit object will be the original covariates with the NAs.

"surr"

Surrogate splitting is used to process NAs. No missingness indicators are created. Nodes are split using only the non-missing values of each variable. To generate predicted values for each unit, a non-missing variable that operates similarly to the variable with missingness is used as a surrogate. Missing values are ignored when calculating balance statistics to choose the optimal tree.

Additional Arguments

Any argument to ps or ps.cont can be passed through weightit or weightitMSM. The argument stop.method is required, and, in contrast to ps and ps.cont, can only be of length 1. If not provided, a default will be used ("es.mean" for binary and multinomial treatments and "p.mean" for continuous treatments).

sampw is ignored because sampling weights are passed using s.weights.

All other arguments take on the defaults of those in ps and ps.cont. The main arguments of interest are n.trees and interaction.depth, and increasing them can improve performance.

Additional Outputs

obj

When include.obj = TRUE, the twang model fit. For binary treatments, the output of the call to twang::ps. For multinomial treatments, a list of ps fit objects. For continuous treatments, the output of the call to ps.cont.

References

Binary treatments

McCaffrey, D. F., Ridgeway, G., & Morral, A. R. (2004). Propensity Score Estimation With Boosted Regression for Evaluating Causal Effects in Observational Studies. Psychological Methods, 9(4), 403<U+2013>425. 10.1037/1082-989X.9.4.403

Multinomial Treatments

McCaffrey, D. F., Griffin, B. A., Almirall, D., Slaughter, M. E., Ramchand, R., & Burgette, L. F. (2013). A Tutorial on Propensity Score Estimation for Multiple Treatments Using Generalized Boosted Models. Statistics in Medicine, 32(19), 3388<U+2013>3414. 10.1002/sim.5753

Continuous treatments

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).

See Also

weightit, weightitMSM, method_gbm

Examples

# NOT RUN {
# Examples take a while to run
# }
# NOT RUN {
library("cobalt")
data("lalonde", package = "cobalt")

#Balancing covariates between treatment groups (binary)
(W1 <- weightit(treat ~ age + educ + married +
                  nodegree + re74, data = lalonde,
                method = "twang", estimand = "ATT",
                stop.method = "es.max"))
summary(W1)
bal.tab(W1)

#Balancing covariates with respect to race (multinomial)
(W2 <- weightit(race ~ age + educ + married +
                  nodegree + re74, data = lalonde,
                method = "twang", estimand = "ATE",
                stop.method = "ks.mean"))
summary(W2)
bal.tab(W2)

#Balancing covariates with respect to re75 (continuous)
(W3 <- weightit(re75 ~ age + educ + married +
                  nodegree + re74, data = lalonde,
                method = "twang", use.kernel = TRUE,
                stop.method = "p.max"))
summary(W3)
bal.tab(W3)
# }