rACME-MEEV

Adjust for Measurement Error in Multiple Exposures from External Validation

How do you adjust for multi-variate measurement error in the absence of an internal validation study? With rACME-MEEV of course! Based off of the work of Muoka et al., 2020 (1), this library takes the proposed methodology and exposes it as a clean R package.

Installation

You can install the development version of rACME-MEEV from GitHub with:

# install.packages("pak")
pak::pak("westford14/rACMEMEEV")

Example

This is a basic example running through a standard workflow of adjusting survey data in the absenece of a validation study.

library(rACMEMEEV)

# Read the data
data <- read_csv("path/to/your/data")
columns <- c("fruit", "vegetable", "tobacco")

# Generate the coefficients
fruit_v_coef <- generate_coefficient(1000, fruit_low, fruit_upper, 0.95)
veg_v_coef <- generate_coefficient(1000, veg_low, veg_upper, 0.95)
tob_v_coef <- generate_coefficient(1000, tobacco_low, tobacco_upper, 0.95)

# Compute the pre-model
output <- acme_model(data, columns)

# Solve the attenuation-contamination matrix
validity_coefficients <- c(fruit_v_coef, veg_v_coef, tob_v_coef)
lambda <- attenuation_matrix(output, columns, validity_coefficients)

# Fit the multivariate model
model_output <- multivariate_model(
  "target ~ fruit + veg + tobacco",
  data = data,
  columns = c("fruit", "veg", "tobacco"),
  jags_model = output$model,
  a_c_matrix = lambda$matrix,
  sds = lambda$sds,
  variances =lambda$variances,
  univariate = TRUE
)

# Plot covariate adjustments
plots <- plot_covariates(model_output, columns)

# Create Traceplots Etc.
acf_plots(model_output$naive, c("fruit", "veg", "tobacco"))
traceplots(output$samples, c("fruit", "veg", "tobacco"))
acf_plots(model_output$naive, c("fruit", "veg", "tobacco"), pre_model = True)
traceplots(output$samples, c("fruit", "veg", "tobacco"), pre_model = True)

Stan

The model has been most thoroughly validated with the JAGS backend, but there is experimental support for the Stan backend. Please do note that it is experimental and liable to change. Here is an example of the same above workflow using the Stan backend.

library(rACMEMEEV)

# Read the data
data <- read_csv("path/to/your/data")
columns <- c("fruit", "vegetable", "tobacco")

# Generate the coefficients
fruit_v_coef <- generate_coefficient(1000, fruit_low, fruit_upper, 0.95)
veg_v_coef <- generate_coefficient(1000, veg_low, veg_upper, 0.95)
tob_v_coef <- generate_coefficient(1000, tobacco_low, tobacco_upper, 0.95)

# Compute the pre-model
output <- acme_model(data, columns, stan = TRUE)

# Solve the attenuation-contamination matrix
validity_coefficients <- c(fruit_v_coef, veg_v_coef, tob_v_coef)
lambda <- attenuation_matrix(output, columns, validity_coefficients, stan = TRUE)

# Fit the multivariate model
model_output <- multivariate_model(
  "target ~ fruit + veg + tobacco",
  data = data,
  columns = c("fruit", "veg", "tobacco"),
  jags_model = output$model,
  a_c_matrix = lambda$matrix,
  sds = lambda$sds,
  variances =lambda$variances,
  univariate = TRUE
)

# Plot covariate adjustments
plots <- plot_covariates(model_output, columns)

# Create Traceplots Etc.
acf_plots(model_output$naive, c("fruit", "veg", "tobacco"), pre_model = TRUE, stan = TRUE)
traceplots(output$samples, c("fruit", "veg", "tobacco"), pre_model = TRUE, stan = TRUE)
acf_plots(model_output$naive, c("fruit", "veg", "tobacco"), stan = TRUE)
traceplots(output$samples, c("fruit", "veg", "tobacco"), stan = TRUE)

Development

If you would like continue development of this library or open up feature requests, you can fork the repo then create a descriptive branch name while continuing devleopment.

git checkout -b your/descriptive/branch/name

Once development is complete, please open a pull request making sure to note the fields provided in the pull request template.

There are a couple of key commands that you should keep in mind when developing this library. First, make sure that devtools and knitr are installed into your R workspace. Then typically the workflow will follow as:

• Make changes to the library
• Run devtools::check()
• Run devtools::install()
• Test the code to make sure it meets your expectations
• Create all necessary new tests
• Run devtools::test()
• Run devtools::document()
• Make any necessary changes to the README.md
• Run knitr::build_readme()
• Finally run make lint from the command line in the root of the repository

Sometimes R will have trouble reading the .db that has all of the namespace exports and will throw an error. So, the easiest way to rectify this is to uninstall and then reinstall the package.

• detach("package:rACMEMEEV", unload=TRUE)
• devtools::install()
• library(rACMEMEEV)

Build Errors

If you are building this package from source there are a few platform specific issues that may arise. This package was developed principally on R version 4.5.1 on an OSX operation system, so RTools may need to be installed on a Windows machine to get proper builds. If there are any issues running devtools::check() that specifically cite RTools, you can try to set the options(buildtools.check = function(action) TRUE ) option so that it skips the explicit local build. This is not recommended, but is a way to speed up the development loop if there are build issues arising.

Code of Conduct

Please note that the rACMEMEEV project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

Maintainers

• westford14

Citations

1. Muoka, A. K., Agogo, G. O., Ngesa, O. O., & Mwambi, H. G. (2020b). A Method to adjust for measurement error in multiple exposure variables measured with correlated errors in the absence of an internal validation study. F1000Research, 9, 1486. https://doi.org/10.12688/f1000research.27892.1