I agree with Lampwick too.
Humans are not a mechanical system to demand from them a 100% repeatability of the results. This repeatability only needs to be more or less distinct, but not absolute. More like a tendency, then a strict law.
For example, in PE people practicing jelqs and stretches in their 1st year of PE, let’s say for the sake of the argument, divide into the following percentiles: 15% see no gains, 40% see an increment of 0,1-0,5”, the next 40% - 0,6-0,8”, the rest 5% - above 0,8” of length gains. If it’s so, then it’s STATISTICALLY SIGNIFICANT.
I.e. you are much more likely to increase the size of your unit by doing PE, then by not doing it (for example, playing computer games instead).
And I dare to state that everything which is statistically significant is scientific ipso facto.
For example, like Lampwick said, the same goes with medicines. Three persons with the same cancer diagnosis use the same medical treatment. One dies abruptly, the second recovers slowly, the third gets healthy very fast. That doesn’t mean that pharmacology is anti-scientific. That simply means there is a quite wide range of possible outcomes and variations in them.
Another example. When a financial minister and a central bank conduct quantitative easing or devalue the currency by 20%, they act on the recommendations of dozens of guys with PhD in Economics and Finance, some of whom even hold Nobel Prizes in Economics. But none of them knows the eventual repercussions of these actions, ‘cause there is a set of possible results. Economics and Finance are even a more imprecise knowledge than PE is.