In r-hyperspec/hySpc.dplyr: Fortify hyperSpec for Use with dplyr

library(hyperSpec)
library(hySpc.dplyr)
library(hySpc.chondro)
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  results = "hide"
)

This vignette introduces package hyperSpec.dplyr.
hySpc.dplyr "fortifies" hyperSpec objects so that they can be used in tidyverse style. In particular, this piping becomes convenient. On a more technical level, hyperSpec.dplyr provides functions such as filter, select, and so on for hyperSpec objects.

The Spectra Matrix

The spectra matrix in $spc behaves a bit different from the extra-data columns, because it contains a whole matrix inside one column. Many tidyverse functions can directly deal with this. In other cases, special attention is needed - this is explained in this vignette.

Selecting a subset of spectra with filter() or slice()

dplyr::filter() selects rows, and filtering by extra data columns works as with any data.frame:

flu %>% 
  filter (c > 0.2)

flu %>% 
  filter (c %>% between (0.15, 0.25))

The filter conditions (logical predicates) must yield one logical value per row. Logical expressions on the spectra matrix such as spc > 100 yield a logical matrix. This matrix indicates which elements of the spectra matrix match the condition.

To obtain the logical vector that indicates which spectra (rows) should be kept, further information is needed: should a spectrum be kept if any of its intensity values (spectra matrix elements) fulfills the condition, or only if all of them do? hyperSpec functions any_wl() and all_wl() summarize the logical matrix in this way.

Typical filtering tasks based on the spectra matrix are:

• remove "empty" spectra which have NAs at all wavelengths (which may be produced by certain import functions): r flu %>% filter (!all_wl (is.na (spc)))
• remove spectra which have NAs at any wavelength (for example after spikes or other artifacts have been marked by NAs): r flu %>% filter (!any_wl (is.na (spc)))
• keep only spectra with all intensities inside a specified range (e.g. non-negative and below a saturation threshold): r flu %>% filter (all_wl (between (spc, 0, 500)))
• keep only spectra with average intensity inside a specified range (e.g. non-negative and below a saturation threshold): r flu %>% filter (rowMeans (spc) %>% between (0, 500)) Here, any_wl() or all_wl() are not needed: the spectra are already summarized into a single intensity value by rowMeans(). The condition therefore evaluates to a one logical value per spectrum already - which is what is needed.

To select rows by indices rather than a condition, use slice():

chondro %>% 
  slice (1:3)

chondro %>% 
  slice (800 : n())

chondro %>% 
  slice (-10 : -n())

flu %>% 
  slice (1, 3, 5)

Also, head() and tail() work as usual.

Selecting particular data columns: select()

select() selects or discards particular columns from a hyperSpec object. If the spectra matrix (colum spc) is included in the selection, the result is still a hyperSpec object:

chondro %>% 
  select (clusters, spc)

If the result does not have the spectra matrix in $spc, it will be returned as data.frame:

flu %>% 
  select (-spc) 

To convert such a data.frame into a hyperSpec object again, use as.hyperSpec:

flu %>% 
  select (-spc) %>%
  as.hyperSpec ()

The resulting hyperSpec object has 0 wavelengths: its $spc column contains a spectra matrix with 0 columns, and its wavelength vector is also of length 0. Behind the scenes, the data.frame returned by select() gets an attribute labels which stores the labels of the hyperSpec object. as.hyperSpec() restores the labels from this attribute (if available). Therfore, the "back-converted" hyperSpec object has its labels preserved.

Renaming extra data columns: rename()

rename() renames extra data columns from the hyperSpec object. Unlike select() rename() keeps all the variables of the data frame intact.

chondro %>% 
  rename(region = clusters)

Mutating data columns for hyperSpec object: mutate()

mutate() adds new variables and preserves all the existing variables.

laser %>%
 mutate (t, filename)
 head # => results in a hyperSpec object

laser %>%
  mutate (-spc) # => results in a hyperSpec object

laser %>%
  mutate (spc2 = spc*2) %>%
  mutate (spc2) %>%
  mutate (spc2*2) # => results in a hyperSpec object

Transmuting data columns for hyperSpec object: transmute()

transmute() mutates the data columns of hyperspec objects. Unlike mutate() transmute()adds new variables but drops existing ones.

laser %>%
 transmute (t, filename)
 head # => results in a data frame

laser %>%
  transmute (-spc) # => results in a hyperSpec object

laser %>%
  transmute (spc2 = spc*2) %>%
  transmute (spc2) %>%
  transmute (spc2*2) # => results in a hyperSpec object

Setting labels: setLabels()

setLabels() Ensure that hyperSpec and non hyperSpec objects have the correct labels.

flu %>%
  setLabels(.wavelength = "f / THz", c = "c / ml")

Selecting wavelength ranges

TODO

---

sessionInfo()

r-hyperspec/hySpc.dplyr documentation built on May 4, 2023, 10:09 p.m.