Skip to contents

Compare different estimators for a (possibly blocked) cluster randomized trial.

Source: R/compare_methods.R
compare_methods.Rd

This function calculates the point estimates and SE estimates for a variety of the estimators implemented by this package.

Usage

compare_methods(
  formula,
  data = NULL,
  control_formula = NULL,
  include_MLM = TRUE,
  include_DB = TRUE,
  include_LM = TRUE,
  include_agg = TRUE,
  include_gee = FALSE,
  warn_missing = TRUE,
  include_dumb = FALSE,
  include_disfavored = FALSE,
  patch_data = TRUE,
  handle_singleton_blocks = c("drop", "pool", "fail"),
  include_method_characteristics = TRUE
)

Arguments

formula

Formula of the form `Yobs ~ Z | clusterID` (or `Yobs ~ Z | clusterID | blockID` if blocked). If NULL, `data` is assumed to already be in canonical form (columns `Yobs`, `Z`, `clusterID`, and optionally `blockID`).

data

Data frame holding the outcome, treatment, and clustering/blocking columns referenced by `formula` (or already in canonical form if `formula` is NULL).

control_formula

What variables to control for, in the form of "~ X1 + X2".

include_MLM

Include MLM estimators

include_DB

Include Design-Based estimators (taken from RCTYes documentation and prior literature).

include_LM

Include Linear Model-based estimators (including Huber-White SEs, etc.)

include_agg

Include estimators applied to aggregated data, aggregated at the cluster level.

include_gee

Include GEE estimators. Off by default: the GEE implementation is still in development and can fail on rank-deficient designs (see `gee_estimators()`).

warn_missing

If TRUE, warn when rows are dropped or imputed due to missing data.

include_dumb

Include "dumb" estimators (i.e., those interacted estimators that weight by both person and cluster or vice versa).

include_disfavored

Include estimators flagged as "disfavored" in `method_characteristics()` (odd weighting or known instability). Also determines whether unrecognized method names are dropped (default) or kept with NA characteristics.

patch_data

If TRUE impute all missing covariates with mean imputation (adding dummy variables as needed) via the `patch_data()` method. Will drop all rows with missing outcome, treatment, or clustering info. If FALSE do not do this.

handle_singleton_blocks

How to handle blocks with a treatment or control arm containing only a single cluster: "drop" the block, "pool" it with another singleton block, or "fail" with an error.

include_method_characteristics

Include details of the methods (target estimands and sampling framework assumed) in the return value.

Value

Dataframe of point estimates and standard errors for each method considered. If include_method_characteristics=TRUE also add some features of the methods as additional columns.

Examples

data( fakeCRT )
compare_methods( Yobs ~ T.x | S.id | D.id, data=fakeCRT )
#>            method  weight   ATE_hat     SE_hat    p_value       df biased
#> 1        AR-FE-db Cluster 0.1852251 0.08740229 0.03537707 189.0000      1
#> 2       AR-FE-het Cluster 0.1852251 0.08735417 0.03527727 189.0000      1
#> 3      AR-FIcw-db Cluster 0.1867940 0.08783710 0.03345371 180.0000      0
#> 4     AR-FIcw-het Cluster 0.1867940 0.08783710 0.03345371 180.0000      0
#> 5      ARpw-FE-db  Person 0.2032865 0.08762329 0.02140859 189.0000      1
#> 6     ARpw-FE-het  Person 0.2032865 0.08660625 0.01994741 189.0000      1
#> 7    ARpw-FIpw-db  Person 0.2044673 0.08567461 0.01700696 180.0000      0
#> 8   ARpw-FIpw-het  Person 0.2044673 0.08600557 0.01743643 180.0000      0
#> 9      LR-FE-crve  Person 0.2032865 0.08660625 0.02015520 157.6603      1
#> 10       LR-FE-db  Person 0.2032865 0.08762329 0.02140859 189.0000      1
#> 11   LR-FIpw-crve  Person 0.2044673 0.08600557 0.01743643 140.2158      0
#> 12     LR-FIpw-db  Person 0.2044673 0.08567461 0.01700696 180.0000      0
#> 13   LRcw-FE-crve Cluster 0.1852251 0.08710924 0.03516777 144.9548      1
#> 14     LRcw-FE-db Cluster 0.1852251 0.08740229 0.03537707 189.0000      1
#> 15 LRcw-FIcw-crve Cluster 0.1867940 0.08780516 0.03338943 134.7822      0
#> 16   LRcw-FIcw-db Cluster 0.1867940 0.08783710 0.03345371 180.0000      0
#> 17         MLM-FE Cluster 0.1931067 0.08674392 0.02720194 186.2165      1
#> 18       MLM-FIcw Cluster 0.1947340 0.08709019 0.02535151 180.0000      1
#> 19         MLM-RE Cluster 0.1816252 0.08666143 0.03744273 187.4170      1
#>    disfavored
#> 1           0
#> 2           0
#> 3           0
#> 4           0
#> 5           0
#> 6           0
#> 7           0
#> 8           0
#> 9           0
#> 10          0
#> 11          0
#> 12          0
#> 13          0
#> 14          0
#> 15          0
#> 16          0
#> 17          0
#> 18          0
#> 19          0