icc9: R Version of the ICC9 SAS Macro

In buddha2490/BERGMets: Metabolomics data and analysis in the BERG

Description

The icc9() function adapts the ICC9 macro published by Ellen Hertzmark and Donna Spiegelman (2010). It computes intraclass correlation coefficients (ICC) and coefficients of variation (CV) and their confidence intervals. These metrics can be calculated with and without adjustment for fixed effects.

Usage

icc9(y,subject,fixed=NULL,dat,method="ML")

Arguments

y	Name of the continuous response vector (REQUIRED)
subject	Name of the subject vector (REQUIRED)
fixed	Vector of fixed effects one or more variable names (OPTIONAL)
dat	Data frame containing variables used in analysis (REQUIRED)
method	Users can indicate "ML" for maximum likelihood method or "REML" for restricted maximum likelihood. By default, the lmer() function calculates estimates based on restricted maximum likelihood (REML). The original SAS ICC9 macro defaults to a maximum likelihood (ML) method, and so that is the default for this R function.

Details

Users familiar with the icc9 SAS macro will notice a few differences with the R function. Notably, the macro options for subsetting the data or including BY= and WHERE= options have been removed. Users should prepare their data prior to running the function.

This function has been tested on a limmited set of data, but has so far replicated the SAS ICC and CV estimates with only an occasional rounding difference. If the user identifies major differences in their dataset, please contact the author of this function with details.

Value

Data frame containing ICC and CV estimates

Response_Variable	Reproduces the response variable for the model (the "y" argument in the function)
N_subjects	Number of unique "subject" values in the input dataset
N_Observations	Number of observations in the input dataset
Average_Measurements	Average number of observations per unique subject
ICC	Intraclass correlation coefficient with confidence intervals
CV	Coefficient of variation with confidence intervals

Author(s)

Brian Carter ([e-mail])

References

Users can refer to the documentation for the original SAS icc9 macro at the following web address: https://cdn1.sph.harvard.edu/wp-content/uploads/sites/271/2012/09/icc9.pdf

Examples

# Example 1 - basic run
icc9(y="X48761", subject="subjid", dat=qc)

# Example 2 - using BATCH1 as a fixed effect
icc9(y="X48761", subject="subjid", fixed=BATCH1, dat=qc)

# Example 3 - Using multiple fixed effects
icc9(y="X48761", subject="subjid", fixed=c(BATCH1,BATCH2), dat=qc)

# Exmample 4 - Using the function for multiple response variables
library(dplyr)
lapply(c("X48761","X19130","X53174"), function(response) {
     icc9(y=response,subject="subjid",dat=qc)
     }) %>% do.call("rbind",.)

buddha2490/BERGMets documentation built on Sept. 6, 2020, 5:11 p.m.