What is meant by Statistical Usage Testing?

Statistical usage testing is the testing process that is aimed at the fitness of the software system or application.

The test cases chosen for carrying out statistical usage testing mostly consist of the usage scenarios and so the testing has been named as statistical usage testing. Software quality is ensured by the extensive testing but that has to be quite efficient. Testing expenditures covers about 20 – 25 percent of the overall cost of the software project. In order to reduce the testing efforts, deploy the available testing tools since they can create automated tests. But usually what happens is that the important tests require manual intervention with the tester requiring thinking about the usage as well behavior of the software. This is just the repetition of the tasks that were done during the requirements analysis phase.

About Statistical Usage Testing

- A usage model forms the basis for the creation of tests in statistical usage testing.

- Usage model is actually a directed usage graph more like a state machine and it consists of various states and transitions. 

- Every transition state has a probability associated with it regarding the traversal of the transition when the system would be in a state that marks the beginning of the transition arc. 

- Therefore, the sum of the probabilities of outgoing transitions sum up to unity for every state.

- Every transition can be associated with an event and more with parameters that are known to trigger the particular transition. 

- Such event associated transitions can be further related to certain conditions called the guard conditions. 

- These conditions imply that the transition occurs only if the value of the event parameter satisfies the condition.

- For assigning probabilities to the transitions, 3 approaches have been defined as follows:

1. Uninformed approach: In this approach, same probability is assigned to the exit arcs of a state.
2. Informed approach: In this approach, a sample of user event sequences for calculating suitable properties. The sample is captured from either an earlier version of the software or its prototype.
3. Intended approach: This approach is used for shifting the focus of the test to certain state transitions and for modeling the hypothetical users.

- According to a property termed as the marcov property, the actual state is what on which the transition probabilities are dependent. 

- However, they are independent of the history again by the property. 

- This implies that the probabilities must be fixed numbers. 

- A system based up on this property is termed as a marcov chain and it requires conclusion of some analytical descriptions. 

- Usage distribution is one among such descriptions. 

- It gives for every state its steady–state probability i.e., appearance rate that is expected.

- All the states are associated to one or the other part of the software system or application and the part of the software that attracts more attention from the tests is shown by the usage distribution. 

- Some other important descriptions are:

1. Expected test case length
2. Number of test cases required for the verification of the desired reliability of the software system or application.

- The idea of the usage model generation can be extended by handling guard conditions and enabling the non–deterministic behavior of the system depending on the state of the system’s data. 

- All this helps towards the application of the statistical usage testing to systems over a wide range. 

- The use cases are defined by the top–level structure of the unified modeling language (UML).