Data

A data object contains your entire n-dimensional dataset, including axes, units, channels, and relevant metadata. Once you have a data object, all of the other capabilities of WrightTools are immediately open to you, including processing, fitting, and plotting tools.

Instantiation

WrightTools aims to provide user-friendly ways of creating data directly from common spectroscopy file formats. Here are the formats currently supported.

name description API
Cary 50 Files from Varian’s Cary® 50 UV-Vis from_Cary50()
COLORS Files from Control Lots Of Research in Spectroscopy from_COLORS()
JASCO Files from JASCO optical spectrometers. from_JASCO()
KENT Files from “ps control” by Kent Meyer from_KENT()
NISE Measure objects from NISE. from_NISE()
PyCMDS Files from PyCMDS. from_PyCMDS()
scope .scope files from ocean optics spectrometers from_scope()
Shimadzu Files from Shimadzu UV-VIS spectrophotometers. from_shimadzu()
Tensor 27 Files from Bruker Tensor 27 FT-IR from_Tensor27()

Is your favorite format missing? It’s easy to add—promise! Check out Contributing.

Got bare numpy arrays and dreaming of data? It is possible to create data objects directly in special circumstances, as shown below.

# import
import numpy as np
import WrightTools as wt
# generate arrays for example
def my_resonance(xi, yi, intensity=1, FWHM=500, x0=7000):
    def single(arr, intensity=intensity, FWHM=FWHM, x0=x0):
        return intensity*(0.5*FWHM)**2/((xi-x0)**2+(0.5*FWHM)**2)
    return single(xi)[:, None] * single(yi)[None, :]
xi = np.linspace(6000, 8000, 75)
yi = np.linspace(6000, 8000, 75)
zi = my_resonance(xi, yi)
# package into data object
axes = []
axes.append(wt.data.Axis(xi, units='wn', name='w1'))
axes.append(wt.data.Axis(yi, units='wn', name='w2'))
channels = []
channels.append(wt.data.Channel(zi, name='resonance'))
data = wt.data.Data(axes, channels, name='example')

Note that channel objects are matrix (ij) indexed. Cartesian (xy) indexed packages like matplotlib will expect the transform.

Structure & properties

So what is a data object anyway? To put it simply, Data is a collection of Axis and Channel objects.

attribute contains
data.axes wt.data.Axis objects
data.channels wt.data.Channel objects

Axis

Axes are the coordinates of the dataset. They have the following key attributes:

attribute description
axis.label LaTeX-formatted label, appropriate for plotting
axis.min coordinates minimum, in current units
axis.max coordinates maximum, in current units
axis.name axis name
axis.points coordinates array, in current units
axis.units current axis units (change with axis.convert)

Axes can also be constants (data.constants), in which case they contain a single value in points. This is crucial for keeping track of low dimensional data within a high dimensional experimental space.

Channel

Channels contain the n-dimensional data itself. They have the following key attributes:

attribute description
channel.label LaTeX-formatted label, appropriate for plotting
channel.mag channel magnitude (furthest deviation from null)
channel.max channel maximum
channel.min channel minimum
channel.name channel name
channel.null channel null (value of zero signal)
channel.signed flag to indicate if channel is signed
channel.values n-dimensional array

Data

As mentioned above, the axes and channels within data can be accessed within the data.axes and data.channels lists. Data also supports natural naming, so axis and channel objects can be accessed directly according to their name. The natural syntax is recommended, as it tends to result in more readable code.

>>> data.axis_names
['w1', 'w2']
>>> data.w2 == data.axes[1]
True
>>> data.channel_names
['signal', 'pyro1', 'pyro2', 'pyro3']
>>> data.pyro2 == data.channels[2]
True

The order of the data.axes list is crucial, as the coordinate arrays must be kept aligned with the shape of the corresponding n-dimensional data arrays.

In contrast, the order of data.channels is arbitrary. However many methods within WrightTools operate on the zero-indexed channel by default. For this reason, you can bring your favorite channel to zero-index using bring_to_front().

At many points throughout WrightTools you will need to refer to a particular axis or channel. In such a case, you can always refer by name (string) or index (integer).

Units aware & interpolation ready

Experiments are taken over all kinds of dynamic range, with all kinds of units. You might wish to take the difference between a UV-VIS scan taken from 400 to 800 nm, 1 nm steps and a different scan taken from 1.75 to 2.00 eV, 1 meV steps. This can be a huge pain! Even if you converted them to the same unit system, you would still have to deal with the different absolute positions of the two coordinate arrays.

WrightTools data objects know all about units, and they implicitly use interpolation to map between different absolute coordinates. Here we list some of the capabilities that are enabled by this behavior.

method description
divide() divide one channel by another, interpolating the divisor
heal() use interpolation to guess the value of NaNs within a channel
join() join together multiple data objects, accounting for dimensionality and overlap
map_axis() re-map axis coordinates
offset() offset one axis based on another
subtract() subtract one channel from another, interpolating the subtrahend

Dimensionality without the cursing

Working with multidimensional data can be intimidating. What axis am I looking at again? Where am I in the other axis? Is this slice unusual, or do they all look like that?

WrightTools tries to make multi-dimensional data easy to work with. The following methods deal directly with dimensionality manipulation.

method description
chop() chop data into a list of lower dimensional data
collapse() destroy one dimension of data using a mathematical strategy
split() split data at a series of coordinates, without reducing dimensionality
transpose() change the order of data axes

WrightTools seamlessly handles dimensionality throughout. Artists and Fit are places where dimensionality is addressed explicitly.

Processing without the pain

There are many common data processing operations in spectroscopy. WrightTools endeavors to make these operations easy. A selection of important methods follows.

method description
clip() clip values outside of a given range
dOD() transform into dOD units
level() level the edge of data along a certain axis
m() apply m-factor corrections [1]
normalize() normalize a channel such that mag –> 1 and null –> 0
revert() revert the data object to an earlier state
scale() apply a scaling to a channel, such as square root or log
smooth() smooth a channel via convolution with a n-dimensional Kaiser window
trim() remove outliers via a statistical test
zoom() zoom a channel using spline interpolation
[1]Absorption and Coherent Interference Effects in Multiply Resonant Four-Wave Mixing Spectroscopy Roger J. Carlson, and John C. Wright Applied Spectroscopy 1989 43, 1195–1208 doi:10.1366/0003702894203408