Design Considerations for Using 1000 Valid and 1000 Invalid Requests in the Experiment

[TedLau]

Dear Authors,
I am very interested in the experimental design in your paper(Specification and Automated Analysis of Inter-Parameter Dependencies in Web APIs). I noticed that when comparing the random testing method and IDLReasoner, you generated 2000 requests for each API, including 1000 valid requests and 1000 invalid requests. I would like to understand the rationale behind this design more deeply:

1.Why did you choose an equal number of valid and invalid requests? Is there any special significance to this balance?
2.What was the main purpose of using invalid requests? How do they help in evaluating the performance of the APIs or the testing methods?
3.How did this design influence your assessment of the performance of the random method versus IDLReasoner?
4.Did you consider using different ratios of valid to invalid requests? If so, why did you ultimately choose the 50/50 split?

Thank you very much for your time and answers. Your response will help me better understand the methodology of this research.