joostscheffer / uncertainty-propagator

An ipynb widget for propagating and evaluating the uncertainty when evaluating a function.

⚠️ This widget is still work in progress ⚠️

Error propagation widget

An ipynb widget for propagating and evaluating the uncertainty when evaluating a function.

Usages

First import the Propagator class from the propagator module.

from uncertainty_propagator import Propagator
import numpy as np

Basic usage

prop = Propagator("I_0 * cos(theta + theta_0) + I_background")

Then use either the widget or the set_variables method to set the variables and their uncertainties. Use the following methods to evaluate the function and the uncertainty at one or multiple points.

y = prop.evaluate_function()
y_err = prop.evaluate_error_function()

Usage with predefined variables

1.  from uncertainties import ufloat as uf
   
    x, a, b = sp.symbols("x a b")
    func = a * x + b
   
    prop = Propagator(func, {b: uf(20, 0), x: uf(10, 1)})

2.  prop = Propagator("I_0 * cos(theta + theta_0) + I_background")
   
    theta = np.linspace(0, 2 * np.pi, 100)
    theta_err = 0.01 * theta
   
    prop.set_variables({
        "I_0": (1.0, 0.1),
        "theta": (theta, theta_err),
        "theta_0": 0.0,
        "I_background": 0.0
    })

uncertainty-propagator's Issues

Separate UI from backend

The app is very promising but for now only seems only usable in jupyter notebooks. Would it be possible to separate the backend into it's own package (or sub package)? This way it wil also be easier to add for example a QT or command line interfaces.

If I have some spare time I would like to spend some time on these functions (If you want some help)

Using dependant errors

Would it be possible to add dependant variables in the propagator. This can be done by for example using an optional correlation matrix on input (doesn't have to be a full matrix only the dependant variables in the set)

Better documentation

The code is not that hard to understand when you take the time to read the doc-strings but some example scripts would really help new users to get going. Also, you already have these nice doc-strings, why not use Spinx to generate some nice documentation?

Testing testing testing

While I trust your math skills development would be a lot less prone to errors if you use some test suites to test basic functionality.