synthesize: Synthesize data

In simongrund1/robosynth: Robust Synthetic Data Generation with Data-Augmented Multiple Imputation

Synthesize data

Description

Synthesizes data with the data-augmented multiple imputation (DA-MI) approach by Jiang et al. (2021).

Usage

synthesize(models, m = 10, iter = 100, verbose = TRUE)

Arguments

models	specification of the combined synthesis and masking models as returned by combine.models.
m	integer: the number of synthetic data sets.
iter	integer: the number of iterations per synthetic data set.
verbose	logical: if TRUE (the default), the progress is printed to the console.

Details

This function synthesizes data with the data-augmented multiple imputation (DA-MI) approach by Jiang et al. (2021). The synthesis model is specified with combine.models and consists of one or more synthesis models (see synthesis.model) that define the relations between the observed variables and a number of masking models (see masking.model) that define the relations between the original variables and masked copies thereof.

Several functions can be used to process the synthetic data. The extract function can be used to extract the synthetic data from the synthetic data object. The with.robosynth.list function can be used to analyze the synthetic data with conventional statistical methods (e.g., lm, glm etc.).

Value

An object of class robosynth.

Author(s)

Simon Grund, Alexander Robitzsch

References

Jiang, B., Raftery, A. E., Steele, R. J., & Wang, N. (2021). Balancing inferential integrity and disclosure risk via model targeted masking and multiple imputation. Journal of the American Statistical Association. Advance online publication. doi: 10.1080/01621459.2021.1909597

See Also

combine.models, synthesis.model, masking.model, with.robosynth.list,
extract

Examples

# NOTE: The number of syntheses (m) and iterations (iter) used here are
# meant to minimize the run time of the examples and should be set to
# larger values in practice (e.g., m = 10, iter = 100).

# * Example 1: DA-MI

# create of masked copies
sociosexuality <- within(sociosexuality, {
  m_sex <- mask.categorical(sex, probability = .80)
  m_sexpref <- mask.categorical(sexpref, probability = .60)
  m_age <- mask.continuous(age, reliability = .90)
  m_behavior <- mask.continuous(behavior, reliability = .95)
  m_attitude <- mask.continuous(attitude, reliability = .95)
  m_desire <- mask.continuous(desire, reliability = .95)
})

# combine synthesis and masking models
models <- combine.models(

  synthesis.model(sex ~ 1, type = "binary"),
  synthesis.model(sexpref ~ 1 + sex, type = "categorical"),
  synthesis.model(age ~ 1 + sex + sexpref, type = "continuous"),
  synthesis.model(behavior ~ 1 + sex + sexpref + age, type = "continuous"),
  synthesis.model(attitude ~ 1 + sex + sexpref + age + behavior,
                  type = "continuous"),
  synthesis.model(desire ~ 1 + sex + sexpref + age + behavior + attitude,
                  type = "continuous"),

  masking.model(m_sex ~ sex, type = "binary"),
  masking.model(m_sexpref ~ sexpref, type = "categorical"),
  masking.model(m_age ~ age, type = "continuous"),
  masking.model(m_behavior ~ behavior, type = "continuous"),
  masking.model(m_attitude ~ attitude, type = "continuous"),
  masking.model(m_desire ~ desire, type = "continuous"),

  data = sociosexuality

)

# run synthesis
syn <- synthesize(models = models, m = 5, iter = 5)

# * Example 2: MI (conventional synthesis)

# combine synthesis and masking models
models <- combine.models(

  synthesis.model(sex ~ 1, type = "binary"),
  synthesis.model(sexpref ~ 1 + sex, type = "categorical"),
  synthesis.model(age ~ 1 + sex + sexpref, type = "continuous",
                  proposal = 5.0),
  synthesis.model(behavior ~ 1 + sex + sexpref + age, type = "continuous",
                  proposal = 2.0),
  synthesis.model(attitude ~ 1 + sex + sexpref + age + behavior,
                  type = "continuous", proposal = 2.0),
  synthesis.model(desire ~ 1 + sex + sexpref + age + behavior + attitude,
                  type = "continuous", proposal = 2.0),

  data = sociosexuality

)

# run synthesis
syn <- synthesize(models = models, m = 5, iter = 5)

simongrund1/robosynth documentation built on March 20, 2022, 6:15 p.m.