Soliciting tests

[glwagner]

I'm going to put together some tests using pytest.

It's fairly simple: we'll just have a folder called /test, and files named test_* that implement tests according to the pytest style.

I'll put in basic unit tests that ensure that critical functions runs, and some "regression" style tests that ensure output is reproducible.

However, we will also need tests that ensure behavior is correct. These are more difficult to come up with. We can discuss ideas on this issue. A few ideas:

• known solutions to the master equation model for simple networks
• statistical tests that make sure the samplers work as intended

[lubo93]

One possible test for the fraction of active edges is to make a comparison with the analytical solution: mean fraction of active edges = lambda(t)/( lambda(t) + mu ).

[glwagner]

@lubo93 and for this test we run the ContactSimulator with no temporal modulation, correct? This should be easy.

[glwagner]

To generate simple networks for tests we can use the following code courtesy of Dr. Boettcher (@lubo93 ):

import networkx as nx

# seed random number generators in *both* numpy.random and random

nodes = 500
attachment = 2

contact_network = nx.barabasi_albert_graph(nodes, attachment)

This generates a graph using the Barbasi-Albert model.