Re: 95% confident
A while back edge.org surveyed a bunch of scientists on the matter of what commonplace scientific idea needed to be discarded. (They've done similar surveys in the past, some collected into books like What is Your Dangerous Idea?, which at the very least make for a fun read.)
Someone - haven't gone back to find out who it was - said "p < 0.05". Using "hey, only one in twenty of this vast catalog of studies is statistically likely to be wrong" as a yardstick for good research is pretty poor.
Of course, when it comes to high-energy physics, the community generally doesn't stop at p < 0.05; that's a very preliminary result. But in a lot of disciplines people are willing to take p < 0.05 as gospel as long as it agrees with their preferences.
(Also of course, the Frequentist-versus-Baysian debate is rather more nuanced than that xkcd strip suggests - Randall's making a joke, even if it has a grain of truth - and lots of frequentist-inclined statisticians deplore that sort of "p < 0.05 therefore I conclude" sloppiness.)