| 2013-03-11 11:03:02 |
Lars Butler |
description |
Adapt the classical hazard qa tests (https://github.com/gem/oq-engine/tree/7d5d79142ead05603118478b46e4e6bbc135f36e/qa_tests/hazard/classical), cases 2-11 to work with the event-based calculator.
In order to get results too converge with expected values (more or less) perfectly, the number of stochastic events needs to be extremely high (100,000s or millions).
The keep the SES count low, we can adopt a variable precision approach in testing against expected values. Something like the following:
For high probabilities, use 2 digits of precision.
For low probabilities, use 3 or 4 digits of precision. |
Adapt the classical hazard qa tests (https://github.com/gem/oq-engine/tree/7d5d79142ead05603118478b46e4e6bbc135f36e/qa_tests/hazard/classical), cases 2-11 to work with the event-based calculator.
In order to get results to converge with expected values (more or less) perfectly, the number of stochastic events needs to be extremely high (100,000s or millions).
The keep the SES count low, we can adopt a variable precision approach in testing against expected values. Something like the following:
For high probabilities, use 2 digits of precision.
For low probabilities, use 3 or 4 digits of precision. |
|
