A basic feature we should keep mind is that we are dealing with random events to which, contrarily to the deterministic ones, the judgments we find out are not accomplished with all certitude but, on contrary are intrinsicality uncertain, i.e., chance-dependent, and inconclusive in what concerns its complete totality. The never-reached final picture should not be limited by what the numbers/observations are directly saying but must be rationally scrutinized and inserted in the context from where they were obtained/detected.
A rational pit-fall
Common sense, which is classic, Pythagorean, leads that to a repeated negative conclusion straighten this negativeness. Surprisedly this so evident result is not followed by Statistics Theory. Suppose that a statistical significance test is performed and the p-value is 10%, repeating result p-value whatever less than 10%. What result by the compound Fisher p - values formula is that globally the set of these two experiments, ?illogically? points towards 5% significance, then leading to a *conventional* Null Hypothesis rejection level.