FactorHet-class {FactorHet}R Documentation

Generic methods for FactorHet models

Description

Brief descriptions of generic methods (e.g. print, summary) for FactorHet as well as a way to visualize the progress of the model-based optimization.

Usage

## S3 method for class 'FactorHet'
plot(x, y = NULL, ...)

## S3 method for class 'FactorHet'
formula(x, ...)

## S3 method for class 'FactorHet'
print(x, fusion.tolerance = 0.001, ...)

## S3 method for class 'FactorHet'
summary(object, show_interactions = FALSE, digits = 3, ...)

## S3 method for class 'FactorHet'
coef(object, coef_type = "beta", ...)

## S3 method for class 'FactorHet'
logLik(object, type = "loglik", ...)

## S3 method for class 'FactorHet'
BIC(object, ...)

## S3 method for class 'FactorHet'
AIC(object, ...)

## S3 method for class 'FactorHet_vis'
print(x, ...)

visualize_MBO(object)

posterior_FactorHet(object)

## S3 method for class 'FactorHet'
vcov(object, phi = TRUE, se.method = NULL, ...)

Arguments

x

Model from FactorHet

y

Not used; required to maintain compatibility.

...

Optional arguments; only used by plot.FactorHet with cjoint_plot.

fusion.tolerance

Threshold at which to declare levels fused

object

Object fit using FactorHet or FactorHet_mbo.

show_interactions

Used by summary.FactorHet; indicates whether the interaction terms be shown. Default FALSE. See "Details" for more discussion.

digits

Number of digits to include

coef_type

Type of coefficient (beta for treatment effects; phi for moderators)

type

For "logLik", should the log-likelihood ("loglik"), log-posterior ("log_posterior"), or sequence of log-posterior values at each iteration ("log_posterior_seq") be returned?

phi

A logical value indicating whether the standard errors from the moderator parameters, \phi, should be returned as well. The default is TRUE.

se.method

A string value for the type of standard errors to be computed. The default, and primary option, is NULL which is generally equivalent to "louis" (Louis 1982), as discussed in Goplerud et al. (2025).

Details

The following methods with the arguments given above exist. All methods work on models with using FactorHet and FactorHet_mbo.

plot:

This is a shorthand for cjoint_plot on a fitted object.

formula:

This returns the underlying formula for the treatment effects and moderators as a named list. This also returns the values used for group, task, and choice_order if provided.

print:

This consists of two print methods. For FactorHet, it summarizes the model and fusion of the factor levels. fusion.tolerance sets the threshold at which levels are reported as fused. For outputs of AME (and similar), this plots the corresponding plot. See that documentation for more details.

summary:

This summarizes the main effects by group with standard errors. It is typically more common to visualize this with cjoint_plot (and the accompanying data.frame) or AME. show_interactions = TRUE shows the interactions in addition to the main effects.

coef:

This returns the coefficient matrix on the original scale (i.e. with the sum-to-zero constraints). code_type = "phi" returns the moderator coefficients instead of the treatment effect coefficients.

AIC and BIC:

This returns the AIC or BIC. If multiple degrees of freedom options specified, it returns a matrix.

logLik:

This returns the log-likelihood, log-posterior or sequence of log-posterior values at each iteration of the algorithm. The argument "type" provides more details.

visualize_MBO:

For a model fit with FactorHet_mbo, this shows information about the MBO, i.e. proposed values and objectives.

posterior_FactorHet:

For a model with K > 1, this visualizes the posterior for each observation and the posterior predictive implied by the moderators.

vcov.FactorHet

This extracts the estimated variance-covariance matrix of the parameters.

Value

Returns the corresponding output of the generic method. "Details" provides details on the output of each function.

References

Louis, Thomas A. 1982. "Finding the Observed Information Matrix when Using the EM Algorithm." Journal of the Royal Statistical Society. Series B (Methodological). 44(2):226-233.

Goplerud, Max, Kosuke Imai, and Nicole E. Pashley. 2025. "Estimating Heterogeneous Causal Effects of High-Dimensional Treatments: Application to Conjoint Analysis." arxiv preprint: https://arxiv.org/abs/2201.01357


[Package FactorHet version 1.0.0 Index]