patient_list: Process patient time series data

In MrMaximumMax/FBCanalysis: Develop and evaluate time series data models based on fluctuation based clustering

Process patient time series data

Description

Process patient time series data by interpolation options and store data in an object of type list.

Usage

patient_list(path, GitHub)

Arguments

path	Path where csv file(s) are stored (only folder, not specific file(s))
GitHub	Set TRUE when csv file comes form GitHub (FALSE by default); only in demo needed

Details

It should be outlined that no prior z-normalization is required! Within FBCanalysis' functions, the user can indicate if normalization is required or not.

Prior to undertaking an analysis using one of the FBC procedures, it is necessary to adequately process and prepare the relevant time series data. The function then creates an interactive flow using the console in R Studio. To begin, the method retrieves all csv files in the provided folder, indicating that it is capable of handling multiple files. The function extracts all csv files from the given directory and merges them into a single raw data frame. The user then indicates which column represents Patient ID and time for adequate processing. The csv files are merged, columns are selected where the Patient ID column will be renamed ”Patient_ID” and the time column will be titled ”Time”. This standardization approach is critical for subsequent features because it enables the easy detection of time series data and the consistent computation and processing of data, for example z-normalization.

The user should also indicate in the interactive console the time formate which will be standardized with the help of lubridate. This is crucial because the technique can now filter the raw data by Patient ID, extract the start and end timestamps for each Patient ID, and then align the data if any records are missing while maintaining the indicated sample frequency.

The user may choose between seven approaches: L1 Regularization/Least absolute shrinkage and selection operator (LASSO) Regression, L2 Regularization/Ridge Regression, Elastic Net Regularization, Linear interpolation, Cubic C2 interpolation or, according to recent articles, fill in missing values using the highest or lowest quartile of measurements in the given time series data distribution.

The Regression and Regularization techniques generate adequate polynomials for each possible degree n-1 (where n is the total number of data points). Afterwards, cross-validation (from glmnet) is applied to determine the lambda value for the lowest MSE of the model. Afterwards, the model with polynomial degree for the lowest MSE is chosen and the missing data is interpolated with the regularized model.

It is also possible to apply a simple linear interpolation in between missing time series data points. It may be the easiest option to employ straight lines between neighboring points (also see na_interpolation). Nevertheless, these basic spline polynomials may be notoriously inexact. Cubic spline polynomials mostly provide better results.

Another option for the user is to apply the interpolation by using a cubic C spline. It implies that the composite function S must be twice continuously differentiable from all boundaries or subintervals (also see na_interpolation).

Without regression, regularization or interpolation, the user may opt to sample missing values within time series data by randomly choosing a value from the greatest or lowest quartile readings from each patient distribution. The R function then loops over each NA element in the time series data distribution of a patient for the specified parameter and randomly samples a value for the chosen quartile until the data frame is complete.

Value

Object of type list storing patient time series data

References

Jerome Friedman, Trevor Hastie, Rob Tibshirani, Balasubramanian Narasimhan, Ken- neth Tay, Noah Simon, and Junyang Qian. Package ‘glmnet’. Journal of Statistical Software. 2010a, 33(1), 2021.

Hui Zou and Trevor Hastie. Regularization and variable selection via the elastic net. Journal of the royal statistical society: series B (statistical methodology), 67(2):301– 320, 2005.

Steffen Moritz and Thomas Bartz-Beielstein. imputets: time series missing value imputation in r. R J., 9(1):207, 2017.

Examples

list <- patient_list(
"https://raw.githubusercontent.com/MrMaximumMax/FBCanalysis/master/demo/phys/data.csv",
GitHub = TRUE)
#Sampling frequency is supposed to be daily

MrMaximumMax/FBCanalysis documentation built on June 23, 2022, 8:21 p.m.