OK , here I go again!
In standard statistical hypothesis testing, to be considered "statistically significant" results need to fall within either a 95% or a 90% confidence interval (CI). What that means is that the likelihood of achieving the same results merely by chance is either 5% (100% -95%) or 10% (100% - 90%), respectively. By convention those are the only CI's used and whichever one is selected is always specified. A paper would never appear in a reputable journal if it wasn't. Both standards are very conservative. No reputable scientist would seriously fault a study for using 90% rather than 95%, especially since it largely depends on sample size. A former professor of mine explained all this to us using baseball data. When comparing runs scored vs. number of wins for one of the leagues using a common statistical test there was found to be no relationship at all at either of the CI's. Only when data from the other league was added increasing the sample size was a significant relationship found but only at the lesser 90% standard. Let's face it. Anyone can tell you that the more runs you score the better chance you have of winning the game. This is elementary, right? It proves that teams who score more runs tend to win more games. Nope, it doesn't. At least not to the same folks who exploit the naivety of the public by discrediting perfectly good science in order to push their agenda. If you think that folks aren't really that ignorant about science and easy to mislead just remember the OJ Simpson trial when Johnny Cochran convinced a jury that DNA evidence isn't really evidence at all! So they let the friggin' guy go free! What's ironic is that people are condemned to death on waaaaay less evidence than something as irrefutable as DNA. Another irony is that scientists, the people best equipped to defend their discipline from those who would pervert it for their own gains, remain largely silent because the minute they speak up they're accused of bias and pushing their own agenda! Nobody wants to damage their reputation so its dammed if they do, dammed if they don't. Imagine if lobbyists and activists of all kinds, journalists, pundits, Political Action Committee members, environmentalists, industry groups, advocates, lawyers, judges, and of course, politicians had some sort of specified standard to live up to, something even remotely resembling the statistical rigor that science demands of itself? By the way, do helmet laws save lives, or does secondhand smoke cause lung cancer, or does fossil fuel combustion warm the planet? Is their any scientific evidence of any of this? It doesn't matter because an industry rep, or lobbyist with a sexy assistant, a box of cuban cigars, and Super Bowl tickets has much more influence on the legislative process than something as capricious and fickle as science.
Walt
