expData: Expanded dataset

In jmpsteen/medflex: Flexible Mediation Analysis Using Natural Effect Models

Expanded dataset

Description

Expanded dataset including either ratio-of-mediator probability weights or imputed nested counterfactual outcomes.

Value

A data frame, resulting from applying neWeight or neImpute on an original dataset data. This data frame has nRep * length(data) rows, containing all original variables (except the original exposure variable) and two variables reflecting observed and hypothetical values of the exposure for each observation unit.

These auxiliary variables (x and x*) are named after the exposure variable and carry integers as suffixes. Suffixes 0 and 1 are used for variables whose corresponding parameters in the final natural effect model index natural direct and indirect effects, respectively.

This object also stores some additional attributes, which are used as input for neModel, such as

model	the fitted working model object
data	original dataset
call	the matched call
terms	the neTerms (internal class) object used
weights	ratio-of-mediator probability weights (only stored if object inherits from class weightData)

Note

If the weighting-based approach (neWeight) is applied, the original outcome values are copied for the nested counterfactual outcomes and the object stores an additional attribute, "weights", containing a vector with ratio-of-mediator probability weights.

If the imputation-based approach (neImpute) is applied, the nested counterfactual outcomes are imputed by predictions from the imputation model.

In the former case, this object inherits from classes c("data.frame", "expData", "impData"), whereas in the latter case it inherits from classes c("data.frame", "expData", "weightData").

See Also

neImpute, neWeight

jmpsteen/medflex documentation built on July 6, 2023, 8 p.m.