Quote:
Originally posted by edstrelow
INon-results are suspect in science. In statistical testing in biological and psychological research, someone who finds no difference hasn't found anything, since that does not justify a conclusion about the non-existence of an effect. |
Well, yes and no. If you have a great deal of statistical power to detect an effect, and you don't, then you actually can say something intelligent about the null hypothesis.
Also, I wouldn't go so far as to say non-results are "suspect" in science. In fact, I'd say it's more the case that *results* are suspect until they've been reliably replicated. (Cold fusion, anyone?)
Quote:
Generally such research will not be accepted for publishing since a non-result could show bad technical skills, and anyone can do an experiment which doesn't work. However, showing an effect beyond the specified sampling error rate or whatever criterion is set up in various disciplines will allow the reseach to be published. |
Well, that's only part of the story. You can beat the statistical criterion and still get papers rejected for poor methodology, bad exposition, or poor theory/argumentation. There's a lot more to it than just showing a low p-value (or whatever).
I think that's what many critics of the green pen thing are reacting to--the notion that the "theory" or explanation behind the green-pen effect is not particularly compelling. Because they do not find this explanation compelling, critics want to see some kind of empirical evidence to the contrary. If that evidence is lacking, the proper "scientific" thing to do is be more conservative, and to go with "no effect."
Quote:
I have seen reports about the "normal" error rate with cd's which showed dozens if not hundreds of errors per second. |
I find this highly unlikely, since CD drives in every computer I've ever used read files with a mean rate of zero errors. It would be obvious if it were not so, since a read error will show up as a reversed bit in a file 50% of the time, and I've never seen it happen except with bad media. Would anyone buy a CD drive if it produced files that were not exact copies of the ones on the CD even 2% of the time per large file, much less dozens of times per second? I think not.
I have a hard time believing that even a mid-fi audio CD reading device would be less accurate at reading bits off a disc than a cheapo computer CD drive, but I accept that it's possible, so I'd really like to see these reports--do you have sources for them?
Quote:
I can't imagine anyone arguing that the sound of cd players hasn't been improved over the last 20 years. Ergo, tweaks can work and are needed. |
I think one could argue quite reasonably that the improvement has been entirely in the DAC stage, and not in accurately reading the bits off the CD. If that's the case, any tweak aimed at somehow improving the "reading the data off the CD" process is suspect. Note that I'm not saying that is the case, I'm just pointing out that the fact that CD players have improved does *not* logically imply that tweaks like the green pen work.