Authors: Bernhard Pollner & Zoltan Kovacs

aquap2 can be useful in three distinct areas in NIRS:

• Experiment Design: Outline the class-structure and generate a truly randomized sample list.
• Data Acquisition: While recording spectra, the previously generated sample list can be used, thus eliminating errors in assigning classes.
• Data Analysis: aquap2 provides an array of standard chemometric procedures as well as some specialized methods used in Aquaphotomics.

To install the R-package "aquap2", do a github-install via devtools, in that way all dependencies should get installed as well.

If you work on a Windows-machine, first download and install R-tools. (Re-start the R-session.)

In the further steps, pay attention to the output of the console.

If not already done, install the package "devtools" via

install.packages("devtools")

Install aquap2 from this repository:

library(devtools)
install_github(repo="bpollner/aquap2", ref="latestPublic")

You can now launch aquap2 and perform (only required at first use) the required setup

library(aquap2)
updateSettings()
updateSettings()

All required code in a single block:

install.packages("devtools")
library(devtools)
install_github(repo="bpollner/aquap2", ref="latestPublic")
library(aquap2)
updateSettings()
updateSettings()

As of October 2023, the dependency from the package ChemometricsWithR has been removed. This should make installation more straightforward, and it is foreshadowing one of the many improvements we have planned for the immediate future.

The repository aquap2_courseMaterial contains a demo dataset with some exploratory and quantitative data analysis steps designed to give a first impression of some of the functions of aquap2. This is the same demo-dataset as handed out at The 3rd Aquaphotomics Symposium in Japan in December 2018.

The course material for the Summer School on Aquaphotomics in the course of the The 3rd European Aquaphotomics Symposium (September 2023, Rome) can be found here.

Please note that the package in its current state is in several parts still considered to be beta:

• From the cross-validated classifier methods, only "fda" and "lda" have undergone a certain amount of testing, the other methods (nnet etc.) have been built in, but still require testing and the implementation of optimization options