Comment Re:I am not reading that. (Score 1) 246
Both from a technical academic nature, and a pure R&D perspective this is very weak argumentation. I'll put it this way: "weak confidence" as you call it never passes the cut. Applying silly non-concrete statistics is more expensive than actually doing the real complete test runs with all possible parameters in many cases. Additionally small sample sizes like this are horribly biased in many cases. Since the technical ones are more iffy (the reasons are often obscure measurement errors), so let me give you a psychology example: We all know the psychology major handing out flyers on campus for a survey, he promises a particular gift (e.g. a can of redbull) and hands out these flyers in a few very specific locations (e.g. faculty of electrical engineering) Then your entire sample group will be highly educated early 20 year olds, and maybe some university faculty. So if you combine that with a small sample size of about 50 you get the most biased world view possible. The same with technical datasets, your small samples will usually be from a single manual production run which means environmental conditions and operator skill and error have a huge impact on the end result.