I am testing a product (liquid bag) for leakage and my primary variable is the amount of leakage in ml. I have a specification that states, on average, leakage must be below XXml. In total I tested 20 bags. Now my problem is this: none of the bags leaked so essentially my data consists of 20 values all equal to zero. Since I have no variability in the data, a t-test offers no utility (average CI for leakage 0 +/- 0 ml????). Aside from treating the data as attribute (leakage/no leakage) and getting a confidence interval for the proportion leaked, does anyone have any opinion on what type of test I perform if I want to keep the data variable (and not as attribute)?
