Wrong results when nan values occur in inflow driven calculations

[cardosan]

Hi Stephan,
when there are nan value in the inflow array it gives wrong results (see below).
converting all the nan to 0 in the stock by cohort array this is solved, at least for the inflow driven calculations

TestDSM = DynamicStockModel(t =  np.arange(1,11,1),
                            i =  [np.nan,np.nan,np.nan,np.nan,np.nan,1,2,3,4,np.nan], 
                            lt = {'Type': 'Normal', 'Mean': np.array([5]), 'StdDev': np.array([1.5]) })
​
In [5]:

Stock_by_cohort, ExitFlag = TestDSM.compute_s_c_inflow_driven()
print(Stock_by_cohort)
[[        nan  0.          0.          0.          0.          0.          0.
   0.          0.          0.        ]
 [        nan         nan  0.          0.          0.          0.          0.
   0.          0.          0.        ]
 [        nan         nan         nan  0.          0.          0.          0.
   0.          0.          0.        ]
 [        nan         nan         nan         nan  0.          0.          0.
   0.          0.          0.        ]
 [        nan         nan         nan         nan         nan  0.          0.
   0.          0.          0.        ]
 [        nan         nan         nan         nan         nan  1.          0.
   0.          0.          0.        ]
 [        nan         nan         nan         nan         nan  0.99240268
   2.          0.          0.          0.        ]
 [        nan         nan         nan         nan         nan  0.9564087
   1.98480535  3.          0.          0.        ]
 [        nan         nan         nan         nan         nan  0.84706865
   1.9128174   2.97720803  4.          0.        ]
 [        nan         nan         nan         nan         nan  0.63410331
   1.6941373   2.86922609  3.9696107          nan]]
In [6]:

S, ExitFlag   = TestDSM.compute_stock_total()
print(S)
[ nan  nan  nan  nan  nan  nan  nan  nan  nan  nan]

[stefanpauliuk]

Hi Cardosan,

thanks for checking this! Indeed, the dynamic stock model class uses summation and does not check whether any or the inputs is NaN, in which case also NaN is returned as result.

Currently I am of the opinion that there should be no such check as the user of the class is responsible for inserting proper numbers. Compare, for example np.dot, which also accepts NaN values.
Will hear with some Python friends about what to do.

Thanks again!

Stefan

[cardosan]

Hey Stephan,
I also actually dunno what is the best way to follow but also if you think that this check should not be done would be good to specify that nan values should not appear in the inflow array in the documentation.

Like I was using the class to calculate stocks for several wood products and countries coming from international statistics and several prod either are not consumed in some countries or their production started only recently and it gave me quite some trouble to discover that this was the reason why the results were very strange.
Cheers
Giuseppe (my real name :) )