README.md

In thomasgredig/RAWdataR: Scientifically Reproducible RAW Data Processing

RAWdataR

RAWdataR provides a framework in R language to uniquely track and analyse raw data from scientific instruments. It assigns unique file IDs based on MD5 sums. This means that even after renaming the files or moving them to different folders, the analysis package can refer to its ID to retrieve the data. Aligned with the National Science Foundation, the Open Data at NSF describes the underlying goals; it also fosters maintaining metadata.

The RAWdataR R package supports the following workflow:

• Creation of unique IDs for all files in a data project
• Basic validation and organization of the scientific data

Installation

You can install the released version of RAWdataR by:

# install.packages("devtools")
devtools::install_github("thomasgredig/RAWdataR")

Documentation

The reference documentation has examples and a list of all functions published in this package.

RAW Data Management Protocol

Raw data is considered the direct output of a scientific instrument (XRD, AFM, etc.). We will follow a few principles for the data R package:

1) The data is stored in a non-proprietory format, such as comma separated values (.csv) and R Data (.rda) or SQLite database (.sqlite); therefore, the data can be accessed beyond the instrument’s lifetime. If an instrument uses a proprietary data format, then it needs to be converted into an open-source data format.

2) All Raw data files in the project are given a unique RAW ID. The ID refers to a particular file and is associated with additional parameters, such as a sample name, temperature, etc. - the ID is limited to the data package and different packages may have the same ID.

3) Multiple collaborators can add data without corruption; i.e. data can be appended without the need for full access to the Raw data.

4) Some Raw data may not become part of the data package to limit the size. Large data files are stored in SQL format (sqlite) outside the package.

5) The data package should include a sufficient description of samples and procedures to understand the instrument RAW data.

Data and Project

The analysis and graphing of the data is in a separate project package. The data package is distinct and provides access to the data in a convenient format. If possible, it does not apply modeling or analysis to the data, but rather makes all data available in a central package.

Data package and Project package play distinct roles.

Naming Convention

In order to achieve scientifically reproducible data, we shall follow these follow principles:

• Raw data cannot be altered and are assigned a unique ID for reliable data analysis.
• Raw data file names can be changed or corrected or stored in a different subfolder without affecting the unique IE, which is automatically reconciled using the MD5 CRC string.
• All data files must be in non-proprietary formats, if not, then a second file with the converted text or ASCII format content needs to be saved as well. If possible, data file names should include the following formatting information:

Date_Project_Initials_Tool_Sample_RunInfo.csv

• sub folders in the RAW folder are useful and can contain additional information; but note that the path file name is mutable and the unique ID must be used.

The date is in yyyymmdd format and represents the date of the data collection start. The project string is assigned by the project manager and the initials are from the person collecting data.

Tools are short strings and represent the machine taking the data, see Tool List.

Each sample should have a unique name, generally starting with the initials of the person, and following the date of sample creation. If more than 1 data collection is made in one day, then RunInfo is added to distinguish or to add more description to the RAW data file.

Example

After you create a new data package, load the RAWdataR package to initialize the package.

library(RAWdataR)
raw.init()

Loading RAW Data

For graphing and data analysis the correct files need to be loaded. A common approach would be searching data files by project, date, user, or by instrument.

# for instance, find all VSM files from 2018
file.list = raw.findFiles(path.RAW, date='2018', instrument='vsm')

As more data is stored, the file.list may change overtime. Therefore, the approach to ensure reproducibility requires the generation of a MD5 string using raw.getPartialMD5str, once the exact file list is established, it can be hard-coded as a string (see below for 4 files). Even if more files are generated, the file list is restricted by the MD5 codes.

md5String = raw.getPartialMD5str(file.list)
file.list = raw.findFiles(path.RAW, date='2018', instrument='vsm',
    md5 = 'a25f3a,66c5d1,4a0333,1b94b5')

Invalid Files

You can also find files with invalid naming convention using the following function, where date is optional

raw.getInvalidFiles(path.RAW, date='2020')

File Checking

Instead of using direct filenames, you can use checksums from the files. For a project that has data added all the time, you could have the following code:

library(RAWdataR)
s = raw.getPartialMD5("README.md")

file.list = dir()
# this will return 'README.md'
filename = raw.getFilename(file.list,s)

Data Project

If you have a new project with RAW data, you can quickly initialize it using the raw.dataProject() function. It will prompt you to select a directory that has a RAW subfolder and then generate the R data package. Once generated, you can open the new data package and run the _init.R code, which helps you prepare a documented data package. Afterwards, it is recommended to add tests to verify data content.

RAWdataR::raw.dataProject()

You can use raw.dataMaker() to generate a make.data..R file that creates a dataset.

Data Standards

We distinguish between digital data from experimental instruments (AFM, XRD, spectrometer, etc.) and analog data (observations, sample names, etc.). All raw instrument data is stored in standardized, open-accessible formats and/or CSV tables, if possible.

A quality control procedure should be in place.

Data analysis will be organized, reproducible and open access. Raw and m’’;’lkjhgfd eta data is saved in R packages. The primary analysis of scientific data is either in R packages or jupyter notebooks. Naming conventions are to be followed.

thomasgredig/RAWdataR documentation built on Nov. 6, 2024, 9:46 a.m.