Peaks, valleys and target values
Four stats are available for peaks and valleys, with the same formal
parameters. These stats do not fit peaks, simply search for local maxima
and local minima in the data as supplied. Stats
stat_peaks() and stat_valleys() subset the
original data while stat_label_peaks() and
stat_label_valleys() only set a boolean flag to mark the
local extremes. Two stats are available for highlighting arbitrary
locations in spectra, one of them, stat_find_wls() accepts
a target for the spectral quantity and locates the corresponding
wavelength values while the other, stat_find_qtys() accepts
a target wavelength value and locates the corresponding spectral
quantity value. Stats stat_find_wls() and
stat_find_qtys() subset the original data or generate new
data by interpolation.
All six stats set the same default aesthetics based on calculated
values. Not all of these default aesthetics are used by the default
geom, but they make using other geoms easier. Furthermore generated text
labels are formatted with sptintf() and these six stats
accept format definitions through parameters label.fmt,
x.label.fmt, and y.label.fmt. Please see the
documentation for a list of all the computed varaibles returned in
data. These stats use internally functions
photobiology::find_peaks() and
photobiology::find_wls() and arguments are passed down to
them.
The examples that follow, apply with minimal changes to
stat_peaks(), stat_valleys(),
stat_find_wls() and stat_find_qtys().
ggplot(sun.spct) + geom_line() + stat_peaks(color = "red")
Because of the conversion, the location of maxima and minima when an
irradiance or response spectrum is expressed in photon- vs. energy-based
units may differ. This is expected and not a bug.
ggplot(sun.spct, unit.out = "photon") + geom_line() + stat_peaks(color = "red")
The complement to stat_peaks() is
stat_valleys.
ggplot(sun.spct) + geom_line() + stat_valleys(color = "blue")
Stats stat_find_wls() and stat_find_qtys()
are another complementary pair. The default target, used in this
example, is the half maximum.
ggplot(yellow_gel.spct) + geom_line() + stat_find_wls(color = "orange")
One of the values calculated and mapped is colour as seen by humans
corresponding to the wavelength at the location of peaks and valleys.
The colour is mapped to the fill aesthetic, so using a
‘filled’ shape results in a colourful plot. The identity scale
is needed so that the correct colours are displayed using the colour
definitions in the calculated data instead of a palette.
ggplot(sun.spct) + geom_line() +
stat_peaks(shape = 21, color = "black") + scale_fill_identity()
We can use any of the aesthetics affecting the default geom,
"point", and also with other _geom_s.
ggplot(sun.spct) + geom_line() +
stat_peaks(span = 35, shape = 4, color = "red", size = 2) +
stat_peaks(span = 35, color = "red", geom = "rug", sides = "b")
We can use several other geoms as needed, demonstrated here
with geom "text". To displace the text we can use
nudging, justification or both. If we want to have the bottom edge of
the label 0.01 y-data units above its natural position we can
use.
ggplot(sun.spct) + geom_line() +
stat_peaks(geom = "text",
span = 35,
color = "red",
vjust = "bottom",
position = position_nudge(y = 0.01))
The same stat can be included more than once in the same plot, using
different geoms. We here in addition demonstrate the use of
several different parameters. The span argument determines
the number of consecutive observations tested when searching for a local
extreme, and it should be an odd integer number. In addition we here
demonstrate the use of a geom new to ggplot2 2.0.0
called "label" which again results in colourful labels by
default. Here we make use of the computed variable BW.color
which is set to "white" or "white" for maximum
contrast with the computed variable fill.
ggplot(sun.spct) + geom_line() +
stat_peaks(shape = 21,
span = 35,
size = 2) +
stat_label_peaks(geom = "label",
span = 35,
vjust = "bottom",
size = 3,
position = position_nudge(y = 0.01)) +
scale_fill_identity() +
scale_color_identity() +
expand_limits(y = 0.9)
Using a larger number as argument to span reduces the number of peaks
detected.
ggplot(sun.spct) + geom_line() +
stat_peaks(shape = 21,
span = 35,
size = 2) +
stat_label_peaks(geom = "label",
span = 35,
size = 3,
na.rm = TRUE,
vjust = "bottom",
position = position_nudge(y = 0.01)) +
scale_fill_identity() +
scale_color_identity() +
expand_limits(y = 0.9)
Setting span to NULL results in the span
set to the range of the data, and so in this case the stat returns the
global extreme, instead of a local one. In this case we use geoms
"vline" and "hline" taking advantage that
suitable aesthetics are set by the stats.
ggplot(sun.spct) + geom_line() +
stat_peaks(span = NULL, geom = "vline", linetype = "dotted", color = "red") +
stat_peaks(span = NULL, geom = "hline", linetype = "dotted", color = "red")
By default the label aesthetic is mapped to a calculated
label x.label giving the wavelength in nanometres. This
mapping can be changed to give to a label giving the y-value at
the peak. We cannot pass nudge_y directly, we need to the
nudge as a position.
ggplot(sun.spct) + geom_line() +
stat_peaks(shape = 21, span = 35, size = 2) +
stat_label_peaks(aes(label = stat(y.label)),
span = 35, geom = "label", size = 3,
position = position_nudge(y = 0.04),
label.fmt = "%1.2f") +
expand_limits(y = 1) +
scale_fill_identity() + scale_color_identity()
ggplot(sun.spct) + geom_line() +
stat_valleys(shape = 21,
span = 35,
size = 2) +
stat_label_valleys(geom = "label",
span = 35,
size = 3,
na.rm = TRUE,
vjust = "top",
position = position_nudge(y = -0.01)) +
scale_fill_identity() +
scale_color_identity()
Above, using geom_label() there is some overlap. We can
use geom_label_repel() from package ‘ggrepel’ to avoid it.
These geometry has additional parameters to which we need to pass
arguments to get a satisfactory positioning of labels.
ggplot(sun.spct) + geom_line() +
stat_peaks(shape = 21, span = 35, size = 2) +
stat_label_peaks(segment.colour = "black",
span = 35, geom = "label_repel", size = 3,
max.overlaps = Inf,
position = position_nudge_repel(y = 0.12),
min.segment.length = 0,
box.padding = 0.5,
force_pull = 0) +
expand_limits(y = 1) +
scale_fill_identity() +
scale_color_identity()
ggplot(sun.spct) + geom_line() +
stat_valleys(shape = 21, span = 35, size = 2) +
stat_label_valleys(segment.colour = "black",
span = 35, geom = "label_repel", size = 3,
max.overlaps = Inf,
position = position_nudge_repel(y = -0.12),
min.segment.length = 0,
box.padding = 0.53,
force = 0.5,
force_pull = 1) +
scale_fill_identity() +
scale_color_identity()
As aesthetics can use values computed on-the-fly we can even
use paste() to map a label that combines both values and
adds additional text.
ggplot(sun.spct) + geom_line() +
stat_peaks(span = NULL, color = "red") +
stat_peaks(span = NULL, geom = "text", vjust = -0.5, color = "red",
aes(label = paste(stat(y.label), "at", stat(x.label), "nm"))) +
expand_limits(y = c(NA, 0.9))
Finally we demonstrate that both stats can be simultaneously used.
One can also choose to use different spans as demonstrated here,
resulting in more maxima being marked by points than labelled with
text.
ggplot(sun.spct) + geom_line() +
stat_peaks(span = 21, geom = "point", colour = "red") +
stat_valleys(span = 21, geom = "point", colour = "blue") +
stat_peaks(span = 51, geom = "text", colour = "red",
vjust = -0.3, label.fmt = "%3.0f nm") +
stat_valleys(span = 51, geom = "text", colour = "blue",
vjust = 1.2, label.fmt = "%3.0f nm")
This final example shows a few additional tricks used in this case to
mark and label the maximum of the spectrum. This example also
demonstrates why it is important that these are stats. The peaks are
searched and labels generated once for each group, in this case each
facet.
ggplot(two_suns.spct) + aes(color = spct.idx) +
geom_line() + ylim(NA, 0.9) +
stat_peaks(span = NULL, color = "black") +
stat_peaks(span = NULL, geom = "text", vjust = -0.5, size = 3,
color = "black",
aes(label = paste(stat(y.label), "at", stat(x.label), "nm"))) +
facet_grid(spct.idx~.)
An additional statistics, stat_spikes(), returns the
same computed variables as stat_peaks() and
stat_valleys() but detects only very narrow peaks and
valleys, usually called spikes. They are common in Raman spectra but can
also appear occasionally in any measurement with array spectrometers
when integration times are long or if some pixels in an array detector
are defective (e.g., hot pixels and dead pixels). Usually spikes are
considered “noise” to be removed, but occasionally we may want to
highlight in a plot the spikes, and do this consistently with function
despike() from package ‘photobiology’.
ggplot(white_led.raw_spct, aes(w.length, counts_3)) +
geom_line() +
stat_spikes(color = "red", z.threshold = 8, max.spike.width = 7)
ggplot(despike(white_led.raw_spct, z.threshold = 8, max.spike.width = 7),
aes(w.length, counts_3)) +
geom_line() +
stat_spikes(color = "red", z.threshold = 8, max.spike.width = 7)
