# Introduction¶

In scientific computations it is a quite common pastime to propagate uncertainty.

The Python Package uncertainties does this sport really catchy. But its propagation yields a numeric result only. It’s great if you only need the final value, but what if you have to document the underlying computation?

$\sigma_y=\sqrt{ \sum_{i=1}^m \left(\frac{\partial y}{\partial x_i}\cdot u_i\right)^2}$

Maabara extends uncertainties package and allows you to calculate and document error propagation in one step. Type your equation once, get the result and its uncertainty including calculation specification - ready in Latex markup. Moreover the data module provides functions for estimation of uncertainty, fitting with error bars, comparison with literature values and more.

Note

At its current state of development Maabara does not support correlations.

## A short example¶

Let’s do something prominent to see how it works: Mass–energy equivalence

import maabara as ma   # import maabara package

einstein = ma.uncertainty.Sheet()           # retrieve computation object
einstein.set_equation('m*c**2', 'E')        # set the equation and name
einstein.set_value('m', 1., 0.1)            # set mass value and its uncertainty
einstein.set_value('c', 28E3, tex='{c_0}')  # set light speed and its Latex symbol

einstein.print_result()

In default setting you’ll get result, equation and propagation of uncertainty like

$E=\left(7.8 \pm 0.8\right) \times 10^{8}$$E={c_0}^{2} \cdot m$$\sigma_{E}={c_0}^{2} \cdot \sigma_{m}$

Of course everything will be copyable Latex markup. Well, it is not too impressing. But add some trigonometric functions, calculate a hundreds of values or generate Latex tables from result and it will get more exciting. Continue reading the User Guide for such pleasures.