Package 'holi'

Launch Shiny App for likelihoodAsy rstar Analysis

Description

This function launches a Shiny application that facilitates the setup and execution of likelihoodAsy rstar analysis. The app allows users to upload a dataset, specify a model and parameters of interest, and perform the analysis with the option to compute confidence intervals for r* statistics.

Usage

LA_rstar()

Value

A Shiny app object that can be run locally.

References

Pierce, D. A., & Bellio, R. (2017). Modern Likelihood-Frequentist Inference. International Statistical Review / Revue Internationale de Statistique, 85(3), 519–541. doi:10.1111/insr.12232

Bellio R, Pierce D (2020). likelihoodAsy: Functions for Likelihood Asymptotics. R package version 0.51, https://CRAN.R-project.org/package=likelihoodAsy.

Examples

if (interactive()) {
  LA_rstar()
}

---

Compute r* Statistics for Generalized Linear Models

Description

The rstar_glm function computes r* statistics for hypothesis testing on coefficients of interest in generalized linear models (GLMs). It supports logistic, linear, and Poisson regression models. For logistic models, the outcome must be binary.

Usage

rstar_glm(
  .formula,
  .data,
  .model = c("logistic", "linear", "poisson"),
  .psidesc = "Coefficient of Interest",
  .psival = 0,
  .fpsi = 2,
  .rstar.ci = FALSE,
  trace = FALSE,
  ...
)

## S3 method for class 'logistic'
rstar_glm(
  .formula,
  .data,
  .model = c("logistic", "linear", "poisson"),
  .psidesc = "Coefficient of Interest",
  .psival = 0,
  .fpsi = 2,
  .rstar.ci = FALSE,
  trace = FALSE,
  ...
)

## S3 method for class 'linear'
rstar_glm(
  .formula,
  .data,
  .model = c("logistic", "linear", "poisson"),
  .psidesc = "Coefficient of Interest",
  .psival = 0,
  .fpsi = 2,
  .rstar.ci = FALSE,
  trace = FALSE,
  ...
)

## S3 method for class 'poisson'
rstar_glm(
  .formula,
  .data,
  .model = c("logistic", "linear", "poisson"),
  .psidesc = "Coefficient of Interest",
  .psival = 0,
  .fpsi = 2,
  .rstar.ci = FALSE,
  trace = FALSE,
  ...
)

## Default S3 method:
rstar_glm(
  .formula,
  .data,
  .model = c("logistic", "linear", "poisson"),
  .psidesc = "Coefficient of Interest",
  .psival = 0,
  .fpsi = 2,
  .rstar.ci = FALSE,
  trace = FALSE,
  ...
)

Arguments

.formula	A formula specifying the model.
.data	A data frame containing the variables in the model.
.model	The type of GLM model: "logistic", "linear", or "poisson".
.psidesc	A description of the parameter of interest.
.psival	The value of the parameter of interest under the null hypothesis.
.fpsi	The index of the parameter of interest.
.rstar.ci	Logical; if TRUE, compute confidence intervals for r*.
trace	Logical; if TRUE, print information about computation. (Default is FALSE)
...	Additional arguments passed to the likelihoodAsy functions.

Value

A list with the object returned from likelihoodAsy::rstar (rs), the object returned from likelihoodAsy::rstar.ci (rs_ci), and the object returned from stats::glm (fit_glm).

References

Pierce, D. A., & Bellio, R. (2017). Modern Likelihood-Frequentist Inference. International Statistical Review / Revue Internationale de Statistique, 85(3), 519–541. doi:10.1111/insr.12232

Bellio R, Pierce D (2020). likelihoodAsy: Functions for Likelihood Asymptotics. R package version 0.51, https://CRAN.R-project.org/package=likelihoodAsy.

Examples

# Logistic model
rstar_glm(law ~ DriversKilled + VanKilled + drivers + kms,
          .data = Seatbelts,
          .model = "logistic") |> suppressWarnings()

# Poisson model
rstar_glm(count ~ spray,
          .data = InsectSprays,
          .model = "poisson") |> suppressWarnings()

# Linear model
rstar_glm(mpg ~ wt + hp,
          .data = mtcars,
          .model = "linear") |> suppressWarnings()

---

Run Multiple Iterations of Simulation and Summarize Results

Description

This function runs multiple iterations of simulation for the sim_rstar_glm function and summarizes the results, including rejection rates, bias, empirical standard error, mean squared error, and root mean squared error.

Usage

run_sim_rstar_glm(
  n_sims,
  alpha_level = 0.05,
  n_main,
  n_covariates,
  true_coef_main,
  n_control = NULL,
  true_coef_control = NULL,
  treatment_effect = NULL,
  model = c("logistic", "linear", "poisson"),
  skewness_main = NULL,
  skewness_control = NULL,
  Sigma_main = NULL,
  Sigma_control = NULL,
  ...
)

Arguments

n_sims	Number of simulations to run.
alpha_level	Significance level for hypothesis tests.
n_main	Number of observations in the main group.
n_covariates	Number of covariates.
true_coef_main	True coefficients for the main group.
n_control	Number of observations in the control group.
true_coef_control	True coefficients for the control group.
treatment_effect	Treatment effect size.
model	Type of model: "logistic", "linear", or "poisson".
skewness_main	Skewness for the main group covariates.
skewness_control	Skewness for the control group covariates.
Sigma_main	Covariance matrix for the main group covariates.
Sigma_control	Covariance matrix for the control group covariates.
...	Additional arguments passed to sim_rstar_glm.

Value

A list with the results of each simulation and a summary of the results.

References

Pierce, D. A., & Bellio, R. (2017). Modern Likelihood-Frequentist Inference. International Statistical Review / Revue Internationale de Statistique, 85(3), 519–541. doi:10.1111/insr.12232

Bellio R, Pierce D (2020). likelihoodAsy: Functions for Likelihood Asymptotics. R package version 0.51, https://CRAN.R-project.org/package=likelihoodAsy.

Examples

sim_summary <- run_sim_rstar_glm(
  n_sims = 2, alpha_level = 0.05,
  n_main = 100, n_covariates = 2, true_coef_main = c(0.5, -0.3),
  n_control = 100, true_coef_control = c(0.2, -0.1),
  treatment_effect = 1, model = "linear"
) |> suppressWarnings()

---

Simulate Data and Fit GLM and r* Models

Description

This function generates simulated data for main and control groups, fits a generalized linear model (GLM) and an r* model, and returns the results.

Usage

sim_rstar_glm(
  n_main,
  n_covariates,
  true_coef_main,
  n_control = NULL,
  true_coef_control = NULL,
  treatment_effect = NULL,
  model = c("logistic", "linear", "poisson"),
  skewness_main = NULL,
  skewness_control = NULL,
  Sigma_main = NULL,
  Sigma_control = NULL,
  ...
)

Arguments

n_main	Number of observations in the main group.
n_covariates	Number of covariates.
true_coef_main	True coefficients for the main group.
n_control	Number of observations in the control group.
true_coef_control	True coefficients for the control group.
treatment_effect	Treatment effect size.
model	Type of model: "logistic", "linear", or "poisson".
skewness_main	Skewness for the main group covariates.
skewness_control	Skewness for the control group covariates.
Sigma_main	Covariance matrix for the main group covariates.
Sigma_control	Covariance matrix for the control group covariates.
...	Additional arguments passed to rstar_glm.

Value

A list with fitted GLM and r* models, and the simulated data.

References

Pierce, D. A., & Bellio, R. (2017). Modern Likelihood-Frequentist Inference. International Statistical Review / Revue Internationale de Statistique, 85(3), 519–541. doi:10.1111/insr.12232

Bellio R, Pierce D (2020). likelihoodAsy: Functions for Likelihood Asymptotics. R package version 0.51, https://CRAN.R-project.org/package=likelihoodAsy.

Examples

sim_result <- sim_rstar_glm(
  n_main = 100, n_covariates = 2, true_coef_main = c(0.5, -0.3),
  n_control = 100, true_coef_control = c(0.2, -0.1),
  treatment_effect = 0.5, model = "logistic"
) |> suppressWarnings()

---

Shiny App for Running r* GLM Simulations with PostgreSQL Integration

Description

This function launches a Shiny application for setting up and running simulations based on the rstar_glm function. The app allows users to input parameters for the simulation, run the simulation, view results, and save results to a PostgreSQL database.

Usage

sim_rstar_glm_pgsql(dbname, datatable, host, port, user, password)

Arguments

dbname	The name of the PostgreSQL database.
datatable	The name of the table in the PostgreSQL database to save the results.
host	The host of the PostgreSQL database.
port	The port of the PostgreSQL database.
user	The username for accessing the PostgreSQL database.
password	The password for accessing the PostgreSQL database.

Value

A Shiny app object that can be run locally.

References

Pierce, D. A., & Bellio, R. (2017). Modern Likelihood-Frequentist Inference. International Statistical Review / Revue Internationale de Statistique, 85(3), 519–541. doi:10.1111/insr.12232

Bellio R, Pierce D (2020). likelihoodAsy: Functions for Likelihood Asymptotics. R package version 0.51, https://CRAN.R-project.org/package=likelihoodAsy.

Examples

if (interactive()) {
  sim_rstar_glm_pgsql(
    dbname = "mydb",
    datatable = "simulation_results",
    host = "localhost",
    port = 5432,
    user = "myuser",
    password = "mypassword"
  )
}

---