Skip to main content

Module providing utilities for working with values with uncertainties

Project description

pycertainties

For all functions f\left( ;x_0,x_1,...,x_i\right), where each value x_i has an associated uncertainty \delta x_i, the uncertainty of the function \delta f, is equal to:

\Large \delta f=\sqrt{\sum_{i}{\left(\frac{\partial f}{\partial x_i}\right)}^2(\delta x_i)^2}

This module contains various utilities for working determining, calculating, and working with values with uncertainties. All of the functions and types in the submodules below are also accesible directly from the pycertainties module.

pycertainties.val

The val submodule provides a single type, Val, that can be used to store values that have an associated uncertainty, and perform calculations on those value as if they were any other number.

For example, the value 10.33 ± 0.12 is represented by Val(10.33, 0.12).

The str conversion returns a sensible representation of the Val.

>>> Val(10.33, 0.12)
10.33 ± 0.12
>>> Val(3.21856e-10, 3.24e-12)
(3.22 ± 0.03)e-10
>>> Val(0.02094495456, 9.541774545e-05)
0.020945 ± 0.000095

We can use the usual operators with Val types.

>>> Val(10.33, 0.12) * 4
41.3 ± 0.5
>>> Val(41.32, 0.48) / 4
10.33 ± 0.12
>>> Val(10.33, 0.12) + Val(3, .5)
13.3 ± 0.5
>>> Val(10.33, 0.12) - Val(3, .5)
7.3 ± 0.5
>>> Val(10.33, 0.12) ** 2
107 ± 2
>>> Val(10.33, 0.12).sqrt()
3.21 ± 0.02
>>> Val(10.33, 0.12).sin()
-0.79 ± 0.07
>>> Val(10.33, 0.12).log()
2.335 ± 0.012

pycertainties.calculations

One possible concern with using Vals are accumulated round-off errors. Especially for computing uncertainties of more complex functions, the amount of intermediate steps can be signifigant, and the round-off errors add up. If these errors are a concern, the calculations.calculate(...) function provides a way to first calculate the uncertainty equation of an equation symoblically (using sympy under the hood), and then calculate the value of that uncertainty equation as the last step.

For example:

>>> calculate("x*y + z", x=Val(3, 0.1), y=Val(3, 1), z=4)
13 ± 3
>>> calculate("x*y + z", x=Val(3, 0.1), y=[Val(3, 1), [Val(5, 1)]], z=4)
[13 ± 3, [19 ± 3]]
>>> import numpy as np
>>> np.array(calculate("x*y + z", x=Val(3, 0.1), y=np.array([[Val(3, 1), Val(5e10, 1e10)], [5, 6]]), z=4))
  [[(13 ± 3) (1.5 ± 0.3)e11] 
 [(19.0 ± 0.5) (22.0 ± 0.6)]]

For a modest speed-up the first argument can also be a sympy expression so that the equation string only needs to be parsed once.

>>> from sympy.parsing.sympy_parser import parse_expr
>>> f = parse_expr("x*y + z")
>>> calculate(f, x=Val(3, 0.1), y=[Val(3, 1), [Val(5, 1)]], z=4) 
[13 ± 3, [19 ± 3]]

The uncertainty equation can also be determined without any values to calculate the final result. This expression can then be used in later calculations or converted to a sympy-parseable string or pretty string. This can be done by calling calculations.uncertainty(expr, *variables) where the variables are all symbols that have an associated uncertainty (equivalent to an uncertainty of 0).

>>> df = uncertainty("x*y + z", "x", "y")
>>> repr(df)
sqrt(x**2*δy**2 + y**2*δx**2)
>>> df
   _________________
  ╱  2   2    2   2
╲╱  x ⋅δy  + y ⋅δx

pycertainties.pprinting

This submodule contains two functions that provide easy ways of visualizing results. The pprinting.pprint_uncertainty(...) function takes arguments of the same form as calculations.uncertainty(...). It will pretty-print the original equation as well as its uncertainty equation.

>>> pprint_uncertainty("x*y + z", "x", "y")
f(...) = xâ‹…y + z
             _________________
            ╱  2   2    2   2
δf(...) = ╲╱  x ⋅δy  + y ⋅δx

The pprinting.pprint_calculation(...) function takes arguments of the same form as calculations.calculate(...). It will pretty-print the original equation, its calculated result, the uncertainty equation, and its calculated result.

>>> pprint_calculation("x*y + z", x=Val(3, 0.1), y=Val(3, 3), z=4)
f(...) = xâ‹…y + z
        -> 13.0
             _________________
            ╱  2   2    2   2
δf(...) = ╲╱  x ⋅δy  + y ⋅δx
        -> 9.00499861188218

pycertainties.strings

This submodule provides a single public function, strings.uncertainty_str(...) that will return a proper string representation of a value and its uncertainty. The expression str(Val(x, y)) is equivilent to uncertainty_str(x, y).

>>> uncertainty_str(10.33, 0.12)
10.33 ± 0.12
>>> uncertainty_str(3.21856e-10, 3.24e-12)
(3.22 ± 0.03)e-10
>>> uncertainty_str(0.02094495456, 9.541774545e-05)
0.020945 ± 0.000095

pycertainties.utilities

This submodule provides a few useful functions for working with uncertainties in numpy.

The utilities.to_val_array(...) function converts two numpy arrays of int/float types to a single numpy array of Val type. The utilities.from_val_arary(...) function converts a single numpy array of type Val to two numpy arrays of int/float types.

>>> to_val_array(np.array([10, 100, 80]), np.array([.1, .15, .35]))
[Val(10, 0.1) Val(100, 0.15) Val(80, 0.35)]
>>> from_val_array(np.array([Val(10, 0.1), Val(100, 0.15), Val(80, 0.35)]))
([ 10. 100.  80.], [0.1  0.15 0.35])

The utilities.weighted_average(...) function calculates a weighted average of a list of Val objects using numpy.

>>> weighted_average([Val(5, .1), Val(100,` 30), Val(10, 1), Val(15, .4)])
5.63 ± 0.13`

Note that the large value of 100 does not appreciably contribute to the average.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pycertainties-1.0.2.tar.gz (10.7 kB view hashes)

Uploaded Source

Built Distribution

pycertainties-1.0.2-py38-none-any.whl (11.0 kB view hashes)

Uploaded Python 3.8

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page