95% confidence and science

Tim Ikeda (tikeda@sprintmail.com)
Mon, 12 Oct 1998 23:18:43 -0400

Joseph Mastropaolo writes elsewhere:
[...]
> Scientists don't believe speculations until put to the experimental
> test and proven with statistical significance, at least P<.05.
[...]

Can't say that I've ever heard that rule. Is this recent?

I've fit curves with ">95% confidence" and known that the results
were bogus. I've also done fits with "<95% confidence" that I've
known were correct. In other cases, I've had no idea how to assign
the likelihood of a particular outcome.

And a couple times I forced-cloned a fragment of DNA with incompatible,
HindIII/EcoRI ends into a complementary vector BACKWARDS. What were
the odds of that? Pretty darn slim... But it happened.

Regards,
Tim Ikeda
tikeda@sprintmail.hormel.com (despam address before use)