Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Test lithoxyl/tests/test_stats.py::test_momentacc_norm is dependent on the random seed #11

Open
sturmianseq opened this issue Aug 12, 2021 · 1 comment

Comments

@sturmianseq
Copy link

If adding random.seed(0) to the beginning of test lithoxyl/tests/test_stats.py::test_momentacc_norm, and running pytest lithoxyl/tests/test_stats.py::test_momentacc_norm, you will see this error message:

____________________________________________________________________ test_momentacc_norm ____________________________________________________________________

    def test_momentacc_norm():
        random.seed(0)
        ma = MomentAccumulator()
        for v in [random.gauss(10, 4) for i in range(5000)]:
            ma.add(v)
        _assert_round_cmp(10, abs(ma.mean), mag=1)
        _assert_round_cmp(4, ma.std_dev, mag=1)
>       _assert_round_cmp(0, ma.skewness, mag=1)

lithoxyl/tests/test_stats.py:49: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

a = 0, b = 0.07368173837804089, mag = 1, name = None

    def _assert_round_cmp(a, b, mag=3, name=None):
        thresh = 1.0 / (10 ** mag)
        abs_diff = round(abs(a - b), mag + 1)
        tmpl = 'round-compare failed at %d digits (%f - %f = %f > %f)'
        rel_diff = (2 * abs_diff) / (a + b)
        err_msg = tmpl % (mag, a, b, abs_diff, thresh)
        if name:
            err_msg = '%r %s' % (name, err_msg)
>       assert rel_diff < thresh, err_msg
E       AssertionError: round-compare failed at 1 digits (0.000000 - 0.073682 = 0.070000 > 0.100000)
E       assert 1.9000637482478797 < 0.1

IMHO, the current way to assert that the skewness is close to 0 is not reasonable, because it in fact asserts that 2*round(skewness,2)<0.1*skewness, which is only true when skewness<0.005 and round(skewness,2)=0. However, the skewness of 5000 random samples can often be greater than 0.005.

@sturmianseq
Copy link
Author

Test lithoxyl/tests/test_stats.py::test_acc_random is also dependent on the random seed. It passes when executing the test class together (i.e., running pytest lithoxyl/tests/test_stats.py), but fails when running in isolation (i.e., running pytest lithoxyl/tests/test_stats.py::test_acc_random) with the following failure message:

============================================================================== FAILURES ===============================================================================
___________________________________________________________________________ test_acc_random ___________________________________________________________________________

    def test_acc_random():
        data = test_sets['random.random 0.0-1.0']
    
        qa = ReservoirAccumulator(data)
        capqa = ReservoirAccumulator(data, cap=True)
        p2qa = P2Accumulator(data)
        for acc in (qa, capqa, p2qa):
            for qp, v in acc.get_quantiles():
                if qp > 0:
>                   assert 0.95 < (v / qp) < 1.05
E                   assert (0.010605121375992893 / 0.01) < 1.05

lithoxyl/tests/test_stats.py:106: AssertionError

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant