Lies, damn lies, and a nifty fusion of theory and practice.

The car-friendly guide to fixing human-made greenhouse emissions.

Free Download!
Sample Chapters

Get the real book from
Accurate Stories

Or, get the ebook from
Smashwords

Buy it for your Smartphone, Tablet, Computer, or eBook Reader, from
Smashwords
Barnes & Noble
Diesel eBook Store
Kobo eBook Store
NOOK

Free Download
Sample Chapters

Lies, damn lies, and a nifty fusion of theory and practice.

Scientific credibility seems to be a popular topic right now, with a multi-billion dollar fusion experiment failing to fi re and a US researcher claiming that most published scientific papers are wrong.

An essay in Nature by freelance writer Regina Nuzzo summarises the state of the debate over the “p-value”, a widely misunderstood test of statistical significance. The p-value estimates the probability that a set of measurements came about by chance. If the p-value is less than 0.05, some researchers will say the result is “statistically significant.” If a study in a sociology or medicine comes up with “p” less than 0.01, it's party time!

Nuzzo explains in relatively plain English why statistical significance doesn't “prove” anything. Over-reliance on p-values over the last fifty years or so has produced a deluge of dodgy “science.” A 2005 paper by John Ioannidis stirred up the debate, and it just won't go away.

There's another way of looking at this. Some people describe the social sciences as “descriptive [academic] disciplines.” Which is a polite way of saying they're not science. Whatever we call them, the message is clear: Statistical significance, on its own, means nothing. Any hypothesis that depends solely on statistical testing doesn't deserve to be called a theory, and it should never be used to make predictions about the future. Sadly, politicians, social scientists, and some journalists seem to love pontificating about this or that such and such that "research says" will happen.

We demand more of physicists. We expect them to come up with practical theories we can convert into engineering tools. Which is probably why management and scientists at the USA's 3.5 billion dollar National Ignition Facility (NIF) have been getting grilled, even though they've achieved what looks like an important milestone on the road to practical fusion engineering.

They created, for the first time, a brief burst of nuclear fusion that produced more energy than it consumed. Never mind that the yield was only about one percent of the energy that went into preparing the fuel. When NIF researchers zapped their fuel with high-powered laser beams, the flickering burst of fusion put out more energy than the laser pumped in. No-one's ever done that before. Trouble is, the folks at NIF said in 2009 they'd achieve ignition by 2012. The US taxpayer forked out 3.5 billion on that expectation.

The NIF's theoretical model said it would happen. No-one would ever have known the model wasn't quite right unless they'd done the experiment. Now we know something wasn't right. Which is a good thing. Twenty-second-century fusion engineers will need reliable mathematical models that accurately predict what fusion reactors will do. If physicists can figure out where the NIF went wrong they might come up with a better model. Maybe even one that works.

Even if they can't, that's no disaster. Physics has already produced plenty of theoretical models that actually work.