Attempting to put this suscinctly (and acknowledging I’m repeating some of what people have already said):
According to the Bible, Adam and Eve were created separately.
Also according to the Bible, Adam and Eve were reproductively compatible.
Conclusion:
Separate creation does not preclude reproductive compatibility.
This would appear to indicate that there is not even a theoretical outcome under which @colewd’s null hypothesis (separate creation) could be rejected in favor of the alternative hypothesis (common descent).
This renders this ‘test’ meaningless.
This is akin to setting the null hypothesis for testing whether a six-sided die is ‘fair’ to be “the sample average is somewhere between one and six, inclusive” – there is no way of rejecting this hypothesis, even if all the numbers rolled are all sixes, or all ones.
Unless Bill can demonstrate that a possible outcome exists under which the null hypothesis (separate creation) would be rejected, Bill’s ‘test’ is meaningless.
Separate creation can easily be rejected when you witness chickens laying eggs and those eggs hatching. Reproduction here is the likely direct cause not special creation.
Addendum: also your complete ignorance of hypothesis testing is showing again – under hypothesis testing the likelihood of the alternative hypothesis is not assessed – only the probability of the null hypothesis.
It’s fine as a null hypothesis. It does not do a great job when comparing the same species but it becomes very effective comparing different species that have large genetic differences.
The question is not quibbling over the null hypothesis. The question is how do we account for the vast genetic differences between species when many amino acid substitutions kill proteins. How can evolution navigate this mine field? Were not even talking function here were just trying to figure out how duplicated sequence A becomes sequence B.
Further evidence that Bill has not the slightest idea of the function of a null hypothesis in statistical hypothesis testing. As I said before:
Addendum: given that separate creation does not preclude reproductive compatibility, there is no potential scenario under which we could, even possibly, reject H0. P(H0)=1 a priori therefore, and the null hypothesis must always be accepted. Therefore any test that has separate creation as its null hypothesis is meaningless.
Further addendum: further explaining this for people who know nothing about hypothesis testing.
In a hypothesis test P(‘Data’|H0) is calculated. If this probability is sufficiently low, H0 is rejected in favor of the alternative hypothesis HA. The probability of HA is not calculated, and typically is incalculable (e.g. for testing a six-sided die, H0 is that the average is 3.5, and HA is that it is “not 3.5” – and P(|‘Data’|“not 3.5”) is incalculable, because it encompasses every possibility other than 3.5, and each of those possibilities has a different probability).
How do we account for the vast genetic differences between individuals of the same species, when many amino acid substitutions kill proteins?
Under the assumption that all organisms are (ultimately) mutants of some distant common ancestor, how can all these organisms live with so many non-lethal mutations, when so many mutations none of the ones alive have are lethal? A conundrum…
All watching modern chickens laying eggs tells you is where those immediate offspring came from.
It does not tell you whether the original origins of chickens involved a single common ancestor of all chickens, or millions of individually created chickens that could lead to millions of ancestors.
This all an interesting idea and I appreciate it but it has little to do with hypothesis testing for common ancestry. Testing two identical species are not going to give us the differences that we need to analyze.
When we are testing for common ancestry we need to identify if the cause of the differences can be attributed to reproduction.
The logical null hypothesis is that the differences that allow for reproductive isolation were not the result of reproductive genetic changes but were from the original design of the animal.
Statistical hypothesis testing involves asserting a value for some statistical parameter – this becomes the null hypothesis. The alternative, that the value of this statistical parameter is different from the asserted value, becomes the alternate hypothesis.
Examples of null hypotheses include:
that the average of a (six-sided) die throw is 3.5 (i.e. that it is a “fair die”)
that the difference between the means of two populations is zero (i.e. that they have the same mean)
that the slope of the graph of two statistics is zero (i.e. that they are uncorrelated)
Therefore to do a statistical hypothesis test of common ancestry, you need to find a statistical parameter that would have a certain value if common ancestry were true, and a different one if it was false.
Alternatively, to do a statistical hypothesis test of separate creation, you need to find a statistical parameter that would have a certain value if separate creation were true, and a different one if it was false.
It is not clear that any single statistical parameter is suitable for either alternative. Common ancestry is therefore likely to be tested by an accumulation/consilience of results, rather than by a single hypothesis test.
Bill has suggested no such statistical parameter for either possibility, so all his blather “has little to do with hypothesis testing for common ancestry.”
I have been familiar with the concept of a “null hypothesis” since I was a teenager, in the 1970s. But I confess that until I met Bill, I never knew there was also such a thing as a “null and void hypothesis.” They are in rich supply here.
Hi Tim
I agree with this approach. Can you think of a way to test common ancestry with a statistical approach?
Design or special creation as the null hypothesis is the claim that most the genetic variation we are observing is based on the original design of the species. Common descent is claiming that the differences can be attributed to reproductive variation.
The current theory is that gene duplication and mutation account for the differences in the gene sequences between species.
We can use the best available empirical data concerning mutation rates, frequency of types of mutations, population sizes, selection coefficients etc to estimate if the genetic differences between animals can realistically be attributed to reproductive variation.
This is what Behe and Snokes were trying to model in 2005. Their model was based on bacteria. Other models have been attempted for vertebrates without providing a serious challenge to the null hypothesis proposed here.
