nluetts/noisysignalintegration.jl

NoisySignalIntegration.jl

A tool to determine uncertainty in numeric integrals of noisy x-y data.

NoisySignalIntegration implements a method to determine the uncertainty in numeric integrals of noisy x-y data on the basis of a Monte-Carlo process. It can include uncertainty due to noise, baseline subtraction, and placement in integration bounds. To do this, the integration is repeated many times while the noise of the data, baseline, and integration bounds are varied based on a noise model and user supplied probability distributions.

To view the documentation, click the badge below:

Installation

The package is not yet registered in Julia's general package registry, you have to install it directly from this Github repository.

To install it for your project, enter the package mode in the Julia REPL (press ]) and type:

add https://github.com/nluetts/NoisySignalIntegration.jl

While still in package mode, you can type

test NoisySignalIntegration

to run the package's unit tests.

Getting Started

Check out the documentation to learn how to use the package.

If you don't have a local Julia installation, you can test the package on mybinder.org. Click the badge below and open one of the example notebooks:

Contributing and Support

If you have problems with the package, submit an issue. Feel free to fork the project and open a pull request if you would like to contribute.

noisysignalintegration.jl's Issues

UncertainSpectra use a lot of memory

Perhaps one can create spectral samples lazily?

local baseline should be defined by points weighted by integration bounds

Currently the local baseline starts and ends at the start and end point of the integration window of the particular draw. This can make the local baseline vary more than one would expect from a look at the spectrum. It would be better to weigh the y-values of the baseline by the spectral data using the distributions of the start and end point of the integration window.

Areas in integral draws plot should have alternating color

Subtracting two curves yields MethodError

MethodError: -(::Curve{Float64}, ::Curve{Float64}) is ambiguous. Candidates:
  -(c::NoisySignalIntegration.AbstractCurve, y) in ...
  -(y, c::NoisySignalIntegration.AbstractCurve) in ...
Possible fix, define
  -(::NoisySignalIntegration.AbstractCurve, ::NoisySignalIntegration.AbstractCurve)

Uninformative error message when trying to integrate outside of curve boundaries

This line

NoisySignalIntegration.jl/src/plotting.jl

Line 27 in 95c6533

) && throw(error("At least one integration bound is outside the support region ($(minimum(x)), $(maximum(x)))."))

references x while it should say xs. When the error is triggered, not the intended error message is shown but an uninformative UndefVarError.

Add function to interpolate Curve on uniform grid

something similar to:

c = sort(Curve(dat.wavenumber, dat.absorbance))
δx = minimum(diff(c.x))
x_even = collect(minimum(c.x):δx:maximum(c.x))
interp = LinearInterpolation(c.x, c.y)
full_spectrum = Curve(x_even, interp(x_even))

Allow for pre-processing step

Perhaps add a callback function that runs before the integration in each MC draw.
For example, this would allow to apply a correction to a spectrum for each draw of the MC process.