When doing analysis, it needs some time knowing which parameter more influence on the results (output) and what the percent influences of each parameter.
After leaving the one parameters each time and including all others and then checking the results may help.
This is a pretty vague question, since you don't mention what the problem is. If it is possible to develop some sort of idealized mathematical model that encompasses the main underlying dynamics, you can then vary parameters one by one and then in various combinations to see how the main outputs are affected. This was the basis, for example, in the early days of DNA sequencing when nobody knew how much data to generate for a given project that might have variations in coverage, bias, fragment distribution, etc. A good mathematical model is your key.