Bug#1008301: statsmodels: test_mvn_pdf intermittent FTBFS/autopkgtest fail - rounding error?

[Rebecca N. Palmer]

Package: python3-statsmodels
Version: 0.13.2+dfsg-1
Severity: serious
Forwarded: https://github.com/statsmodels/statsmodels/issues/8195

(The high severity is because FTBFS is RC by policy - I consider it 
likely that this is within normal rounding error and should be ignored.)

test_mvn_pdf is intermittently failing in a high-required-accuracy 
comparison.

=================================== FAILURES 
===================================
_______________________ TestMVDistributions.test_mvn_pdf 
_______________________

self = 
<statsmodels.sandbox.distributions.tests.test_multivariate.TestMVDistributions 
object at 0x7fa1b57c1cf0>

     def test_mvn_pdf(self):
         cov3 = self.cov3
         mvn3 = self.mvn3

         r_val = [-7.667977543898155, -6.917977543898155, 
-5.167977543898155]
         assert_array_almost_equal(mvn3.logpdf(cov3), r_val, decimal=14)
         # decimal 18
         r_val = [0.000467562492721686, 0.000989829804859273,
                  0.005696077243833402]
 >       assert_array_almost_equal(mvn3.pdf(cov3), r_val, decimal=17)
E       AssertionError:
E       Arrays are not almost equal to 17 decimals
E
E       Mismatched elements: 1 / 3 (33.3%)
E       Max absolute difference: 1.56125113e-17
E       Max relative difference: 4.40579422e-15
E        x: array([0.00046756249272168, 0.00098982980485927, 
0.00569607724383339])
E        y: array([0.00046756249272169, 0.00098982980485927, 
0.0056960772438334 ])

/usr/lib/python3/dist-packages/statsmodels/sandbox/distributions/tests/test_multivariate.py:136: 
AssertionError