Create a weightit object manually

This function allows users to get the benefits of a weightit object when using weights not estimated with weightit() or weightitMSM(). These benefits include diagnostics, plots, and direct compatibility with cobalt for assessing balance.

Usage

as.weightit(x, ...)

# S3 method for weightit.fit
as.weightit(x, covs = NULL, ...)

# S3 method for default
as.weightit(
  x,
  treat,
  covs = NULL,
  estimand = NULL,
  s.weights = NULL,
  ps = NULL,
  ...
)

as.weightitMSM(x, ...)

# S3 method for default
as.weightitMSM(
  x,
  treat.list,
  covs.list = NULL,
  estimand = NULL,
  s.weights = NULL,
  ps.list = NULL,
  ...
)

Arguments

x

required; a numeric vector of weights, one for each unit, or a weightit.fit object from weightit.fit().

...

additional arguments. These must be named. They will be included in the output object.

covs

an optional data.frame of covariates. For using WeightIt functions, this is not necessary, but for use with cobalt it is. Note that when using with a weightit.fit object, this should not be the matrix supplied to the covs argument of weightit.fit() unless there are no factor/character variables in it. Ideally this is the original, unprocessed covariate data frame with factor variables included.

treat

a vector of treatment statuses, one for each unit. Required when x is a vector of weights.

estimand

an optional character of length 1 giving the estimand. The text is not checked.

s.weights

an optional numeric vector of sampling weights, one for each unit.

ps

an optional numeric vector of propensity scores, one for each unit.

treat.list

a list of treatment statuses at each time point.

covs.list

an optional list of data.frames of covariates of covariates at each time point. For using WeightIt functions, this is not necessary, but for use with cobalt it is.

ps.list

an optional list of numeric vectors of propensity scores at each time point.

Value

An object of class weightit (for as.weightit()) or weightitMSM (for as.weightitMSM()).

Examples

treat <- rbinom(500, 1, .3)
weights <- rchisq(500, df = 2)

W <- as.weightit(weights, treat = treat, estimand = "ATE")
summary(W)
#>                  Summary of weights
#> 
#> - Weight ranges:
#> 
#>            Min                                   Max
#> treated 0.0123 |---------------------------| 11.9693
#> control 0.0039 |------------------------|    10.6706
#> 
#> - Units with the 5 most extreme weights by group:
#>                                             
#>             451    340    496     25     216
#>  treated 7.6391 7.9751 8.6735 9.3784 11.9693
#>             404    265    260    348      52
#>  control  8.102 8.1265 9.3111 9.4798 10.6706
#> 
#> - Weight statistics:
#> 
#>         Coef of Var   MAD Entropy # Zeros
#> treated       1.023 0.773   0.455       0
#> control       0.984 0.747   0.426       0
#> 
#> - Effective Sample Sizes:
#> 
#>            Control Treated
#> Unweighted  335.    165.  
#> Weighted    170.41   80.91

# See ?weightit.fit for using as.weightit() with a
# weightit.fit object.