I would say that in doing an experiment that will have multiple trials, if one of the variables is changed each time, say temperature or pressure amount, then average results between experiments will not be valid. They would only be valid if the same conditions applied for each one, to determine the inherent variability of the result.