Further to astronomer Hugh Ross on degrees of certainty in science, from Christie Aschwanden, Five Thirty-Eight’s lead science writer:

P-values have taken quite a beating lately. These widely used and commonly misapplied statistics have been blamed for giving a veneer of legitimacy to dodgy study results, encouraging bad research practices and promoting false-positive study results.

But after writing about p-values again and again, and recently issuing a correction on a nearly year-old story over some erroneous information regarding a study’s p-value (which I’d taken from the scientists themselves and their report), I’ve come to think that the most fundamental problem with p-values is that no one can really say what they are. More.

Here’s the theory, from Dummies, but apparently no one finds it easy to understand in practice:

For example, suppose a pizza place claims their delivery times are 30 minutes or less on average but you think it’s more than that. You conduct a hypothesis test because you believe the null hypothesis, Ho, that the mean delivery time is 30 minutes max, is incorrect. Your alternative hypothesis (Ha) is that the mean time is greater than 30 minutes. You randomly sample some delivery times and run the data through the hypothesis test, and your p-value turns out to be 0.001, which is much less than 0.05. In real terms, there is a probability of 0.001 that you will mistakenly reject the pizza place’s claim that their delivery time is less than or equal to 30 minutes. Since typically we are willing to reject the null hypothesis when this probability is less than 0.05, you conclude that the pizza place is wrong; their delivery times are in fact more than 30 minutes on average, and you want to know what they’re gonna do about it! (Of course, you could be wrong by having sampled an unusually high number of late pizza deliveries just by chance.)

*See also:* Nature: Banning p-values not enough to rid science of shoddy statistics

and

Rob Sheldon explains p-value vs. R2 value in research, and why it matters

Oh, and Steven Weinberg defends “Whiggish” history of science Actually, science has nothing over these other endeavours when the question can be decided by evidence.

Follow UD News at Twitter!

Although mathematics and real time testing certainly play a very important part, ‘observation’ is the backbone of science.

This is where Darwinian evolution and Intelligent Design drastically part company. Whereas we know for 100% certainty that intelligence can generate non-trivial functional information, no one has ever observed even one instance of unguided material processes generating non-trivial functional information.

In fact, just one ‘observed’ instance would falsify ID.

In fact, Perry Marshall has organized a 3 million dollar prize for the first person who can prove that unguided material processes can generate non-trivial information:

There are solid mathematical reasons for believing that that 3 million dollar prize will never be collected:

The reason I bring this complete lack of observational evidence for Darwinism up, is because some Darwinists a few years back tried some shenanigans with these statistical p-values to make it appear as if Darwinian evolution were something more than the atheistic pipe dream that it is.

A few more notes on Darwinism’s complete lack of ‘observational’ evidence:

You need to have a fair understanding of what sort of distribution and mean would make sense for a null, or a p-value connected to say a normal curve can be very misleading. Do you have an idea of a very large cluster of small effectively unbiased errors pushing a particular observation this way and that at random around a mean? Try a Galton Board, cf vid: https://www.youtube.com/watch?v=9tTHST1sLV8 here is a professional version: https://www.youtube.com/watch?v=AUSKTk9ENzg If that is not plausible, normal curve circumstances do not obtain, though this is hardly the only possibility. KF

great intuitive view of probability and random chance too

That “For dummies” example is pretty awful. The real definition is not hard to grasp: it’s the probability of getting a data equal to or more extreme than an observed value if a null hypothesis is true.

The problems only arise when we people miss-specify a null hypothesis or conflate this probability with an event happening “by chance”, the probability the null hypothesis is true or the probability some alternative hypothesis is false. (Or whatever Rob Sheldon is going on about in that linked post).

If p-value is a probability term, -log2(p-value) is the information in bit. I wonder if “p-value” can be used to estimate CSI? The consequence is that many statistical tests can then be used to estimate the p term in CSI.

From wiki, “the p-value is defined as the probability of obtaining a result equal to or “more extreme” than what was actually observed, assuming that the hypothesis under consideration is true.”

I think p-value is a very clear and useful concept. It is true that many who use it don’t understand what it is (that is sadly true in medicine!).

And yes, when we evaluate dFSCI (or CSI) we use a form of probability under null hypothesis: more or less, for example, for a functional protein, we ask: what is the probability to get to a protein which exhibits at least a pre-defined level of a pre-defined function if we assume that the observed functional protein came into existence by a random walk from some unrelated sequence?

And then we compare that probability with the probabilistic resources available in a natural system.

Indeed — the last form of CSI that dembski defended was a reformulation of statistical hypothesis testing.