Adam, Eve and Population Genetics: A Reply to Dr. Richard Buggs (Part 1)


(Richard Buggs) #810

Hi all,

Just for completeness I would like to note that Dennis has shared Charles Cole’s new analyses with me, @Swamidass and @glipsnort in a private forum back in February. These analyses do not demonstrate that PSMC could detect a bottleneck of two followed by rapid population growth. Rather, they add evidence that this method could not detect such a bottleneck. I am not sure if Dennis is still intending to share these analyses here on the public forum or not.

My understanding is that @DennisVenema now agrees that PSMC estimates of past effective population sizes could not detect a bottleneck of two followed by rapid population growth.


(George Brooks) #811

@RichardBuggs

Meaning, it did not have the capacity ? Or that it does, and it found no evidence for such?


(Richard Buggs) #812

Of possible interest:

“Evidence collected at sites in the basin suggests that early humans underwent a series of profound changes at some point before roughly 320,000 years ago. They abandoned simple hand axes in favour of smaller and more advanced blades made from obsidian and other materials obtained from distant sources. That shift suggests the early people living there had developed a trade network — evidence of growing sophistication in behaviour. The researchers also found gouges on black and red rocks and minerals, which indicate that early Olorgesailie residents used those materials to create pigments and possibly communicate ideas…The team cannot say exactly how long before 320,000 years these changes happened because an extended period of erosion at the site wiped out the archaeological record there between 499,000 and 320,000 years ago.”


(Richard Buggs) #813

It does not have the capacity. Thanks for asking for clarification. This is my understanding of @DennisVenema’s view. I am sure he will provide an update on his own views soon, as I may be an imperfect messenger.


#814

Sorry, couldn’t help it.


(George Brooks) #815

@RichardBuggs

Thanks for the link to that article!

Great final two paragraphs…

Some information could come from several projects that drilled into ancient lake beds in Kenya and Ethiopia to collect a detailed record of environmental and ecological changes in the region6. Potts and his team drilled two of those cores in the southern Olorgaseilie Basin, and Potts says the cores cover the entire period that is missing from the archaeological record. Comparisons with cores drilled elsewhere in East Africa should help scientists to differentiate between events happening locally and broader regional climatic trends.

“The drill cores I hope will be a game changer, because of the precision of the environmental record and hopefully the precision of the dating,” Potts says. Then it’s a matter of working to understand how animals and people might have responded to the changing environment, Potts says. “Only then can we say anything about how climate is really affecting human evolution.”


(Dr. S Joshua Swamidass) #816

d[quote=“RichardBuggs, post:813, topic:37039, full:true”]
It does not have the capacity. Thanks for asking for clarification. This is my understanding of @DennisVenema’s view. I am sure he will provide an update on his own views soon, as I may be an imperfect messenger.
[/quote]

That is my understanding of the evidence, and also my understanding of how @DennisVenema sees it.

This post is important to go over:

However, PSMC has several considerable limitations that should be kept in mind.

  1. It doesn’t recover sudden changes in Ne
  2. Nor does it recover recent changes, e.g. younger than 10,000 years BP in humans (Li and Durbin 2011).
  3. Simulation suggest that it also performs worse in case of very ancient changes in Ne (Mazet et al. 2015).
  4. Using incorrect mutation rate or generation time can cause bias in the interpretation.
  5. The change in Ne in a PSMC plot can be actually caused by population structure.

Another must read paper is here, Mazet et al 2016


(Jon Garvey) #817

Very relevant, Joshua. The problem with models of the distant past is partly, as these articles relate, the possible problems we know that we know - like possible changes in mutation rate.

But models in the real world are potentially - and often actually - falsified by actual failed outcomes: the medical intervention one models doesn’t actually alter the death rate, so you go back, invstigate, and find factors you didn’t include in the model.

One very tragic and graphic example - we’re told that the accelerated building technique for the bridge at Florida International University hadn’t been used before. But you can bet it was extensively modelled, and found good. But something was missed in the model - as in many other engineering examples. Only reality cruelly uncovers the flaws in the modelling.

In the end, for the past population of humanity we wouldn’t need a population genetics model if we had sufficient information about the deep past to validate it. And it’s vanishingly unlikely that the fossil record contains enough information for us ever to be able to correct any unknown problems with the models from actual data.

“All models are wrong, but some are useful.” That’s very true, but we must always remember what it means: models show us what kind of data to look for in the real world: their usefulness is entirely about their match to reality. Internal consistency is satisfying to the modeler, but not “useful”.


(Chris Falter) #818

Hi Jon,

Hope things are going well on your side of the pond.

When scientists build a model well, they use statistical terminology to define how much uncertainty needs to be acknowledged. Joshua @Swamidass has done a good job of this IMO, and stated IIRC that the TMR4A is 500kya ± 100k.

Do you have any reason to dispute the boundaries of the cone of uncertainty that @Swamidass has placed around his model? If so, how would you quantify the uncertainty so we appreciate it better?

Thanks,
Chris


(Jon Garvey) #819

Chris, it’s snowy here, but we should be able to burst out to the outside world tomorrow.

My point was, essentially, about things we don’t know that we don’t know. That can be checked on a bridge by building a bridge, or on model for widget marketing by marketing widgets.

But factors operating half a million years ago, on an unknown number of individuals in an unknown geographic distribution and even an unknown number of species or subspecies, and with the mechanisms of speciation not fully understood are potentially legion.

If you have a way of quantifying what one doesn’t know that one doesn’t know, you’re a better clairvoyant than I am!


(George Brooks) #820

@Jon_Garvey

Fortunately for @DennisVenema, his methodology is more secure in the first 25,000 years than at any other time.


(Chris Falter) #821

Hi Jon, Now I understand why England athletes excelled at the winter games!

Based on my reading of the scientists’ discussion, I think they have been able to quantify the impact of the unknowns you cite.

As for the unknown unknowns, I leave the last word to my invisible Martian friend, ÷=/_<>××:

//=×>[÷÷_</=÷×66==÷<>>

Very compelling, ÷=//_<>××! Thanks for sharing your perspective!


(Jon Garvey) #822

Ah - “Statistical Analyses are the assurance of things hoped for, the conviction of things not seen.” (Heb 11:1, American Scientistic Version)

“The epidemiologist fallacy is also richer than the ecological: it occurs whenever an epidemiologist says, “X causes Y” but where he never measures X and where he uses classical statistics to claim proof of a cause - based on, say, wee p-values or large Bayes factors; [citation]. Over-certainty is guaranteed.” (William Briggs)

X, in this case, is the original ancestral population that gives rise to the population we have now, which we are trying to ascertain and which we cannot measure, even if we want to. If we really can, through models alone, allow for all the factors that we don’t know we don’t know, then science is back to the Greek model, where reason replaces empirical examination of the world.

Unknown Martians are an unlikely factor here, in my humble view. But unknown biological mechanisms, and unknown historical contingencies, are not only likely but turn up every year - what do we know now from actual data about the radiation of humanity that we did not know this time last year? Wasn’t “Out of Africa” not long ago an uncomplicated and more or less uncontested understanding of where we came from?

One possible confounding factor for this particular discussion, as regards the final truth of the matter, is the ultimate point at issue (correctly excluded by Richard for the scientific discussion) - a known God who, according to many following the traditional understanding of the Scriptures, created the first humans as a single couple de novo.

I have no dog in the fight - I don’t believe an Adam >200K years ago is consistent with Genesis. But I do know that once one has done the excellent modelling that Joshua, in particular, has worked on, then what one has achieved is a fuller exploration of the model(s) currently in use, and a bigger range of possibilities than was previously thought - but by no means a certainty about the deep past on which to stake your life.

Or even a reliable quantification of what surprises the world may have for you.


(Jon) #823

Let me guess, you also dispute anthropogenic climate change.


(Jon Garvey) #824

Guess wrong. The forecasts of climate change can be tested by direct data… not that I’ve checked the records on how they’re doing currently.


(Jon) #825

What happened to all the things that we don’t know that we don’t know?


(Jon Garvey) #826

Jon

I think you missed the point. When real-world data defy the predictions, the model is proven wrong. If the data is unavailable, as it is for the deep past, the agreement of as many independent models as you like does not “prove” a single one of them. They may only suggest probabilities - but the real world’s certainties have a habit of overturning assessments of probability.

I don’t know what your own working experience is in science, but I spent my entire career in clinical medicine. Every intervention was based on models built, to speak crudely, from applying current theory to available data to determine beneficial outcomes. That applied to therapeutics, surgical procedures, epidemiology, or anything else.

Over four decades, I saw possibly a majority of those models falsified, invariably when someone got round to looking more closely at the real-world outcomes. Sometimes that just meant factoring in new variables to tweak the model, but often it meant an entire reversal of previous “evidence-based” practice. That’s no criticism of my profession - except when it was so certain of the models that nobody bothered to test them against reality, of which I could name many examples. In every case you can go and look up the research on which those faulty models were built, and why they were wrong - in retrospect.

Ask any old doctor, and they will agree that, new drugs etc aside, it’s only because people’s bodies are so resilient that we got away with much of what we did in the 1970s.

Ten years after retirement, when I speak to my colleagues it’s clear that, if I returned to work, much of the advice or treatment I gave would have become outmoded: but only by the replacement of the models were were using back then by new ones based on data from the real world forcing a change of theory and, hence, a different model.

Your experience of science may be more ideal - in my world it is messy and provisional.


(Jon) #827

In this case, what data is unavailable?


(Jon) #828

So you would reject the claims of people like Roy Spencer who say the current warming is a natural cycle?


(Christy Hemphill) #829

@Jon Garvey’s personal views on climate change aren’t relevant to the discussion. If you two would like to delve into this topic, do it via PM.