[onewhippedpuppy]

Originally Posted by jyl

How accurately are you trying to measure a given characteristic?

What I mean is, and this probably isn't the right terminology for aerospace, but suppose you are trying to measure fuel consumption at a given airspeed, altitude, etc. Say your tests return sonething around 6.5 gph, making up a number. Do you test enough to determine that it is 6.3 to 6.8 gph to a 95% confidence?

Same question for other performance metrics, rate of climb or roll or whatever.

It totally depends on the test, the more critical the data the tighter the requirements. Performance data is very intensive, where flight conditions can have very tight tolerances on things such as altitude, angle of attack, stabilized airspeed, etc. Landing data is critical too with things like sink rate, angle of attack, and airspeed. Ultimately it's all difficult for the pilot to hit, so lots of tests are required to get enough data. I'm not one of the data crunching types so I don't know what specific statistical confidence interval they work to, but it's pretty tight and all subject to FAA review. There's a reason why airplanes are expensive and take years to certify.