rACMEMEEV (version 1.0.1)

multivariate_model: Model the Data with Multivariate Adjustment

Description

Using the methodology generated by Muoko et. al (see README for full citation), run the modelling with the JAGS sampler. This model uses the calculated attenuation-contamination matrix to adjust the various covariates needed in the model. The function accepts the computed pre-model, the attenuation-contamination coefficient,, a model string, and standard deviations. There is an optional argument for also fitting the univariate model to run further comparisons between the naive, univariate, and multivariate models in later steps.

Usage

multivariate_model(
  formula,
  data,
  columns,
  a_c_matrix,
  n_burn = 1000,
  n_thin = 1,
  n_steps = 10000,
  seed = 42,
  b0 = 0,
  capital_b_0 = 1e-06,
  sampler = "Metropolis",
  c0 = 0.001,
  d0 = 0.001,
  univariate = FALSE,
  sds = NULL,
  variances = NULL
)

Value

List with fitted models ready for further analysis

Arguments

formula

character The model formula

data

data.frame The data to model

columns

vector The columns that are relevant

a_c_matrix

matrix The attenuation-contamination matrix

n_burn

numeric Number of draws to burn at the start, default = 1000

n_thin

numeric Thinning factor for the draws, default = 1

n_steps

numeric The total number of draws to run, default = 10,000

seed

numeric The random seed to set

b0

numeric The prior mean of the beta

capital_b_0

numeric The precision of the beta

sampler

character The sampler to use for the model

c0

numeric Shape parameter for gamma

d0

numeric Scale parameter for gamm

univariate

bool Whether or not to run the univariate model, default = TRUE

sds

list If you are running the univariate model, the listing of standard deviations needs to be applied

variances

list If you are running the univariate model, the listing of variances need to be applied

Examples

columns <- c("fruit", "veg", "tobacco")
fruit_v_coef <- generate_coefficient(100, 0.3, 0.8, 0.95)
veg_v_coef <- generate_coefficient(100, 0.25, 0.75, 0.95)
tob_v_coef <- generate_coefficient(100, 0.4, 0.7, 0.95)
validity_coefficients <- c(fruit_v_coef, veg_v_coef, tob_v_coef)
data <- data.frame(
 list(
   "BMI" = rnorm(100, mean = 0, sd = 1),
   "fruit" = rnorm(100, mean = 0, sd = 1),
   "veg" = rnorm(100, mean = 0, sd = 1),
   "tobacco" = rnorm(100, mean = 0, sd = 1)
 )
)
output <- acme_model(data, columns)
lambda <- attenuation_matrix(
  output,
  columns,
  validity_coefficients,
)
model_output <- multivariate_model(
  "BMI ~ fruit + veg + tobacco",
  data = data,
  columns = columns,
  a_c_matrix = lambda$matrix,
  sds = lambda$sds,
  variances = lambda$variances,
  univariate = TRUE
)