The spectacles
package is making it easy (or at least easier!) to handle spectroscopy data. It provides the user with dedicated classes (namely Spectra
and SpectraDataFrame
), so that most of the useful information about the spectral dataset is available in one R object:
- the spectral values
- the wavelengths at which these have been recorded
- some kind of ID
- if available, some associated data (typically, some lab measurements)
The stable version of spectacles
is on CRAN (:tada:):
install.packages('spectacles')
You can also install the development version using the devtools
package:
# Install devtools if you don't have it on your machine
# install.packages('devtools')
devtools::install_github("pierreroudier/spectacles")
It also provides easy ways to plot a collection of spectra:
- simple line plots of the individual spectra
- offset plots of the individual spectra
- stacked plots of the individual spectra
- summary plots of a whole collection, or aggregated against a given factor
- tools to code more advanced visualisations yourself using eg
ggplot2
orlattice
It also gives overloads to the most common operators such as $
, [
, or [[
, so that any user familiar with data.frame
object would fell right at home.
The philosophy of the package is really just to make it easier to work with quite complex data. There are a lot of tools already existing in R to do spectral preprocessing (signal
, etc.). A few additional tools have been added in spectacles
, such as the ASD splice correction.
The idea is for the package to work quite well with the pipe (%>%
) operator from the magrittr
package, to create chains of pre-processing operators. The function apply_spectra
makes it easy to work with any function whose input is either a numeric
vector or a matrix
:
# Example of splice correction, followed by
# a first derivative, followed by a SNV
my_spectra %>%
splice %>%
apply_spectra(diff, 1) %>%
apply_spectra(snv)
# Another example using prospectr
my_spectra %>%
splice %>%
apply_spectra(prospectr::continuumRemoval, wav = wl(.)) %>%
plot
Again, lots of existing methods available, so spectacles
is not re-implementing any of these. There's various ways to use spectacles
with the different methods available, but my favoured option is to use it in conjonction with the caret
package, which gives a unique API to 160+ models in R:
fit <- train(
y = s$carbon,
x = spectra(s),
method = "pls"
)
spectroSummary(fit)
Yes, I once had a package called inspectr
on Github, and spectacles
is very much the continuation of inspectr
. The only reason why inspectr
changed name is that someone pushed a package called inspectr
on CRAN (despite inspectr
being quite visible on Github.... :-/). So, lesson learnt this time!