rACMEMEEV (version 1.0.1)

sensitivity_analysis: Run a sensitivity analysis on the error adjustment

Description

Create a sensitivity analysis based on a grid of validity coefficient parameters. The input of the parameter space should be a list with each key being the name of one of the covariates to vary and a vector of the parameter space to test the sensitivity. Be very careful because all combinations of the parameters will be tested so you can very easily run this for too long. Also please note that the parameters should be be bound between 0 and 1, the theoretical limits of the validity coefficients. An example of this parameter grid is:

Usage

sensitivity_analysis(
  parameters,
  data,
  formula,
  columns,
  stan = FALSE,
  seed = 42
)

Value

list with the means and the SDs of the parameters

Arguments

parameters

list As described above

data

data.frame The data to use for the sensitivity analysis

formula

character The formula for the model

columns

vector The columns within the data for the covariates

stan

boolean Whether or not to run with the Stan backend

seed

numeric The random seed to set

Details

params <- list( fruit = c(0.1, 0.2, 0.3), veg = c(0.1, 0.2, 0.3), tobacco = c(0.1, 0.2, 0.3) )

But, again please note that this will then fit 3 * 3 * 3 different models so the run time here can explode. Also parallel computation will be utilized here, but it is much more difficult to debug should errors arise. All this to say is: buyer beware.

Examples

if (FALSE) { # identical(tolower(Sys.getenv("NOT_CRAN")), "true")
columns <- c("fruit", "veg", "tobacco")
data <- data.frame(
 list(
   "BMI" = rnorm(5, mean = 0, sd = 1),
   "fruit" = rnorm(5, mean = 0, sd = 1),
   "veg" = rnorm(5, mean = 0, sd = 1),
   "tobacco" = rnorm(5, mean = 0, sd = 1)
 )
)
parameters <- list(
  fruit = c(0.3),
  veg = c(0.25),
  tobacco = c(0.4)
)
output_jags <- sensitivity_analysis(
  parameters,
  data,
  "BMI ~ fruit + veg + tobacco",
  columns
)
}