ReorganizeAndMICE: Preps data and runs Mice

In jordanaron22/ImputingMetabolites: Compares R2 Value of Different Metabolite Imputation Methods

Description

From CreateCorrelation() correlation matrices are in 3d arrays with dimensions num_of_metabo x num_of_metabo x total number of studies First we change the 3d array into a 2d matrix with dimensions (num_of_metabos x (num_of_metabos-1)/2 by total number of studies Each column corresponds to an individual study and each row to a pairwise correlation of metabolites. Then we use the fischer transform and run MICE on the transpose of the prepped 2d matrix. 5 MICE iputations are done and then the median is taken for each pairwise value corr_array_mice is a 3d array with dimensions num_of_metabo x num_of_metabo x total number of studies that has been un-fischer transformed. reorganized_corr_imp_med is a 2d matrix with dimensions (num_of_metabos x (num_of_metabos-1)/2 by total number of studies that is still fischer transformed. reorganized_corr_imp_med will be the input for the next step of the imputation pipeline.

Usage

ReorganizeAndMICE(
  corr_array_obs,
  factor_mat,
  num_of_metabos = 20,
  num_of_predictors = 0
)

Arguments

corr_array_obs	Observed correlation array. Output from CreateCorrelation()
factor_mat	Matrix where 3 columns are three factors and rows are values of each factor
num_of_metabos	Number of metabolites
num_of_predictors	Number of other metabolites for MICE to use, set to 0 to use quickpred()

Value

un-fischer transformed 3d array corr_array_mice and fischer transformed 2d matrix reorganized_corr_imp_med

jordanaron22/ImputingMetabolites documentation built on Dec. 21, 2021, 2:18 a.m.