compare_methods: Compare different estimators for multisite/blocked RCTs.

In lmiratrix/blkvar: ATE and Treatment Variation Estimation for Blocked and Multisite RCTs

Compare different estimators for multisite/blocked RCTs.

Description

This function calculates the point estimates and SE estimates for a variety of blocked designs implemented by this package. It is a wrapper for the many other individual methods available, and returns things in a tidy table.

Usage

compare_methods(
  Yobs,
  Z,
  B,
  siteID = NULL,
  data = NULL,
  include_block = TRUE,
  include_MLM = TRUE,
  include_DB = TRUE,
  include_LM = TRUE,
  include_DBBlended = FALSE,
  include_method_characteristics = FALSE,
  weight_LM_method = "survey",
  weight_LM_scale_weights = TRUE,
  control_formula = NULL
)

Arguments

Yobs	vector observed outcomes (or column name in data)
Z	vector of assignment indicators (1==treated) (or column name in data)
B	vector of block ids (or column name in data)
siteID	site ids (variable name as string if data frame passed) (if randomization blocks are nested in site).
data	frame holding Y, Z, B and (possibly a column with name specified by siteID).
include_block	Include the standard error estimation found in Pashley (in particular, these work for blocks with singleton units).
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_DBBlended	Include DB estimators applied to small block (blocks with singleton treatment or control units) and classic Neyman to large blocks.
include_method_characteristics	Include details of the methods (target estimands and sampling framework assumed) in the return value.
weight_LM_method	Passed to weight.method of weighted_linear_estimators; specifies the weighting method.
weight_LM_scale_weights	This argument is passed to scale.weights of weighted_linear_estimators
control_formula	What variables to control for, in the form of "~ X1 + X2".

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.

lmiratrix/blkvar documentation built on Nov. 18, 2024, 1:27 p.m.