Taxonomy as a Rigorous Science

[Jodie]

In certain places, where applicable. I'm asking you to demonstrate that this is one of those places. Simply throwing out numbers doesn't cut it--you need to show that the numbers are relevant to paleontology, and that not using them causes problems. If not using statistics doesn't cause any problems (and "it seems wrong to Jodie" isn't a problem) then there's no issue.


If I showed you pictures of chimpanzees, what sample size would you need to determine that they were not pictures of humans?

Since when has statistics not cut it? That's what Dinwar says they use, did you not read the opening post and the discussion points in the other thread? My statistical example should highlight exactly why it is subjective. To compare Neanderthal to a chimpanzee is not a realistic comparison. First, we don't have photographs of Neanderthal, second the morphological differences between a chimp and a modern human is pretty marked, not so with some of the other ancient hominids like Neanderthal.

You have 400 individuals represented out of an estimated 70,000. Those 400 span the known epoch that Neanderthal existed, assuming that none are found that are older or newer than what exists in the collection today. These 400 show the gradual changes over time moving from the more archaic version to those sharing similar features with modern man.

According to Dinwar, they measure these features to determine whether they belong in the category, or the species assigned. He also stated that statistics are used to establish those parameters, meaning that they use the frequency of variation that you might find in other species that are similar to establish a mean or average degree of variation. The math works when dealing with a small sample like 400, if you are only looking within the 400 for comparison, using established guidelines for variations across species. But what you don't have based on the estimated population is the 95% confidence level that a larger sample size would give you to account for the variation across the entire species of Neanderthal as opposed to just the 400. In my mind, this means that classification of those with mixed features might actually fall into the wrong category without DNA analysis to back up the decision.

Going back to your example of one specimen- how would you know, without DNA, that it wasn't a member of homo sapiens sapiens? It might be a variation caused by a fluke of genetic problems ( think Star Child skull as a single skull prototype that doesn't necessarily indicate a new species) or a disease process if you didn't have the DNA to establish that it was a different species. Even that is debatable to some since the thought is that there was some co-mingling. So does this gradual appearance of modern features in Neanderthal indicate that modern man also contributed to Neanderthal's DNA?? I don't know, they were either evolving in the same way as modern humans or they weren't as different as we think since co-mingling occurred. I haven't found any genetic research that specifically addresses what modern man might have contributed to the Neanderthals' genome.

 

[kellyb]

Jodie, do you know how to use google scholar?

 

[Jodie]

I didn't know there was such a thing, is it good?

 

[kellyb]

I didn't know there was such a thing, is it good?

It is really really really good. Imagine a pubmed of all the sciences that takes you right to the full texts a lot of times.

http://scholar.google.com/

I have it at the top of my favorites bar so I can quickly locate original research to answer my questions. It also has a nifty "cited by" feature so you can see who cited (and maybe either verified or refuted) the article you just read.

Supercool stuff.

 

[Akri]

Since when has statistics not cut it? That's what Dinwar says they use, did you not read the opening post and the discussion points in the other thread?

We're talking about the sample size needed to say that a specific creature is or is not part of a particular species. Statistics is used in certain parts of paleontology, but not in the way you appear to be arguing.

My statistical example should highlight exactly why it is subjective.

It doesn't. The best you can argue is that conclusions drawn by paleontologists might turn out to be wrong (something they're well aware of), but that is NOT the same as being subjective. Nothing you have shown in any way demonstrates subjectivity on the part of scientists. You have also failed to show any errors which have occurred as a result of using sample sizes that you consider too small.

In my mind, this means that classification of those with mixed features might actually fall into the wrong category without DNA analysis to back up the decision.

Yes, I get that you think this might be a problem. My point is, unless you can demonstrate that it actually is a problem, why should we care? The science appears to be working just fine as-is. Just because you can imagine a particular problem doesn't mean that problem actually exists.

Are neanderthal specimens being mislabeled? Can you actually demonstrate that this is a problem? Or does this exist purely in the realm of "things Jodie imagines might be wrong"?

Going back to your example of one specimen- how would you know, without DNA, that it wasn't a member of homo sapiens sapiens?

Imagining for the purposes of this hypothetical that I was a paleontologist with the skills necessary to identify hominid skulls, I would know based on the identifying features. I'm not a paleontologist so I can only guess at what those are, but they probably have to do with jaw length, number and shape of teeth, brain cavity size, and things like that.

It might be a variation caused by a fluke of genetic problems ( think Star Child skull as a single skull prototype that doesn't necessarily indicate a new species) or a disease process if you didn't have the DNA to establish that it was a different species.

What are the odd of such genetic fluke happening? Of those who have the fluke, how many are likely to end up fossilized?

Let's say that out of 70,000 Neanderthals, 10% have some deformity that makes their skulls look indistinguishable from that of a modern human. So, 7,000 have this defect.

Now, out of all 70,000 Neanderthals, we have a whopping 400 fossils. That's a little over .5%. If we're only getting .5% of the entire population fossilized, then the odds of getting one of those 7,000 fossilized is only .05%.

So out of 70,000 Neanderthals, there might be 7,000 with this defect. Of those 7,000 there might be three or four that get fossilized--provided, of course, that reality matches the statistics perfectly.

Now let's imagine that you're a paleontologist looking at a freshly unearthed Neanderthal skull. Which is more likely: that it's one of the estimated 396 normal skulls, or one of the 4 anomalous skulls?

This will hold true for any population you look at. Fossilization is extremely rare, so the less prevalent a trait is within a population, the less likely that trait is to be fossilized. So while it is possible for a given fossil to contain some incredibly rare anomaly, the odds are very much stacked against it. Thus, unless there's some specific reason to think you're dealing with an anomalous specimen, it's pretty safe, mathematically speaking, to discard that possibility.

 

[marplots]

What are the odd of such genetic fluke happening? Of those who have the fluke, how many are likely to end up fossilized?

Let's say that out of 70,000 Neanderthals, 10% have some deformity that makes their skulls look indistinguishable from that of a modern human. So, 7,000 have this defect.

Now, out of all 70,000 Neanderthals, we have a whopping 400 fossils. That's a little over .5%. If we're only getting .5% of the entire population fossilized, then the odds of getting one of those 7,000 fossilized is only .05%.

So out of 70,000 Neanderthals, there might be 7,000 with this defect. Of those 7,000 there might be three or four that get fossilized--provided, of course, that reality matches the statistics perfectly.

Now let's imagine that you're a paleontologist looking at a freshly unearthed Neanderthal skull. Which is more likely: that it's one of the estimated 396 normal skulls, or one of the 4 anomalous skulls?

How did you go from the 10% (top highlight) to 1% later on?

 

[rockinkt]

Dinwar... My knowledge of your particular branch of science is very limited. Therefore, I appreciate your efforts in putting together such a well reasoned and informative essay.

However, because I'm wearing black socks today - I'm going to disagree with everything you said and call your particular science - "Bollocks".

(This message was brought to you on behalf of the closet bleevers of JREF and certified "the real deal" by Bigfoot Science™)

 

[Akri]

How did you go from the 10% (top highlight) to 1% later on?

10% is the made-up chance of an individual having this hypothetical deformity. .05% is the chance that an individual will have the deformity, AND get fossilized.

Looking at it now, I think the final number is actually .0005, not .05 (.5% of 10%). In which case the math says that out of 400 fossils, .02 will have this fluke. So, none, basically.

Feel free to check my math, though. I'm not that great at math to begin with, and coming off of a two-day migraine has not improved my skills any. (I did run the numbers by someone else who tends to be fairly decent at math, but I think he may have been only half paying attention).

The point, at any rate, is that fossilization is exceedingly rare. A genetic fluke that makes a member of species A look like a member of species B is also pretty rare. So the odds of an individual with that fluke getting fossilized are extremely small, to the point where it doesn't make much sense to worry about them.

 

[marplots]

10% is the made-up chance of an individual having this hypothetical deformity. .05% is the chance that an individual will have the deformity, AND get fossilized.

Looking at it now, I think the final number is actually .0005, not .05 (.5% of 10%). In which case the math says that out of 400 fossils, .02 will have this fluke. So, none, basically.

Feel free to check my math, though. I'm not that great at math to begin with, and coming off of a two-day migraine has not improved my skills any. (I did run the numbers by someone else who tends to be fairly decent at math, but I think he may have been only half paying attention).

The point, at any rate, is that fossilization is exceedingly rare. A genetic fluke that makes a member of species A look like a member of species B is also pretty rare. So the odds of an individual with that fluke getting fossilized are extremely small, to the point where it doesn't make much sense to worry about them.

Well, if that formulation held up (10% of deformity times 1% chance of fossilized - switched to 1% for easy math) you'd have .1 X .01 of getting a deformed fossil for an answer of .001. But then, if you work the numbers the other way, for the normals, you get .9 X .01 for a total of .009

The mistake is that we have switched meaning so that the .001 and the .009 have become the chances of finding a normal vs. an abnormal fossil. But notice, those chances to not total to 100%. They should, if we find any fossils at all.

In reality, the ratio of deformed to normal cannot change, unless there is some reason that fossils will be made from one class rather than another. (Maybe Neanderthals dumped their misfits into swamps or something.) If fossilization is a random selection from the larger pool, the proportions would be the same. What might change with the small number of fossils is the likelihood of variance exceeding some set level - is that what you meant?

In that case, where a small sample means it's easier to move away from the average condition, you would expect to get more samples that don't match the predicted percentages, so there's a point to be made there.

The new question is this then: If I take a sample of 400 from a population of 70,000 where 10% show deformity, what are the chances that none in my sample will show the deformity?

 

[Akri]

Hmm, you may be right, marplots. It's a bit after 2am here, so I'm gonna get some sleep and then look at this tomorrow with fresh eyes.

 

[Jodie]

(62,300/700) / (400/0)

0.0001 odds that all Neanderthals in the 400 sample set will have no genetic abnormalities?

 

[kellyb]

(62,300/700) / (400/0)

0.0001 odds that all Neanderthals in the 400 sample set will have no genetic abnormalities?

When it comes to genetic abnormalities that would make them appear to be a different species, that doesn't seem too far off. That's a different thing from "no genetic abnormalities", though.

 

[Jodie]

Yep, that's why I asked for clarification. Neanderthals do show deformity as a result of environmental impact. I don't think I've read where someone thought they had genetic defects.

Akri asked why it was important to be accurate, I think it's pertinent in understanding what happened before to get an idea of what to expect in the future for mankind. Nothing is ever static, you adapt or die.

 

[Akri]

In reality, the ratio of deformed to normal cannot change, unless there is some reason that fossils will be made from one class rather than another. (Maybe Neanderthals dumped their misfits into swamps or something.) If fossilization is a random selection from the larger pool, the proportions would be the same.

After giving it some thought, this doesn't really make sense.

If I have a bag with 90 red marbles and 10 blue ones, and I ask you to reach in with your eyes closed and grab five, you shouldn't expect those five to have the same proportions as the bag (4 red and 1 blue). You should expect all five to be red.

A better way to look at it is, how big does the sample size need to be for it to be likely that a single individual in the sample (one marble or one fossil) comes from that 10% minority?

 

[Akri]

Akri asked why it was important to be accurate

Citation needed.

 

[kellyb]

Neanderthals do show deformity as a result of environmental impact. I don't think I've read where someone thought they had genetic defects.

There's rather a lot of stuff like this in google scholar:

http://peer.ccsd.cnrs.fr/docs/00/59/12/88/PDF/PEER_stage2_10.1002%2Fhumu.21389.pdf

Ancestral PCMs
considered to be disease-causing in humans were identified in two Neanderthal genes (DUOX2,
MAMLD1). Since the underlying mutations are known to give rise to recessive conditions in
human, it is possible that they may also have been of pathological significance in Neanderthals.

Note the cautious "it is possible" language, too.

I think the more you read the actual research, the more impressed you'll become with the quality/rigor as well as the caution exercised when they're interpreting results from incomplete evidence.

My personal sense is that the paleos/geneticists are overall somewhat better in this regard than a lot of the people publishing on current human disease epidemiology (I have a whole hypothesis on why that might be, but that's a topic for a different thread.)

 

[Sideroxylon]

Very interesting stuff. I'll probably have to read through it a few more times to actually understand it (somewhere after the dotted line my brain started going "bwuh?")

It never occurred to me that taxonomy worked as a test for evolution--that's neat!

In the observation of these nested hierarchies, there is a very powerful and potentially simple to understand argument for evolution in there, especially when tied in to biogeography.

Here is Kent Hovind studiously not understanding it:
http://www.youtube.com/watch?v=NWhCJy0AJgM

 

[marplots]

After giving it some thought, this doesn't really make sense.

If I have a bag with 90 red marbles and 10 blue ones, and I ask you to reach in with your eyes closed and grab five, you shouldn't expect those five to have the same proportions as the bag (4 red and 1 blue). You should expect all five to be red.

A better way to look at it is, how big does the sample size need to be for it to be likely that a single individual in the sample (one marble or one fossil) comes from that 10% minority?

We can figure the odds for your marble example by looking at the odds that all will be red. It's a combination of the probabilities for getting red on each of five separate draws:

90/100 x 89/99 x 88/98 x 87/97 x 86/96 = about 58% of the time. That means the odds of getting at least one blue (could get more), is 42%.

For your original example, it gets a bit more complicated because we are drawing 400. So let's switch to factorial notation. The odds of no deformities showing up at all are:

63,000/70,000 x 62,999/69,999 x 62,998/69,998 ... and so on four hundred times. For the top term, we get 63,000!/(63,000 - 400)! and for the bottom term, we get 70,000!/(70,000 - 400)!

Amazing to me, the calculator in Windows can handle large factorials. Here's what I get for the final answer: 4.38^-19

I'm not sure I have any confidence in this answer and wouldn't mind a double check. Just to get an idea of how it works, I did the first five terms by hand (as if we found just five instead of four hundred).

For that case, the odds of no deformities showing up are 60% and falling rapidly as the number found gets bigger. So, by four hundred, it's going to be pretty small.

 

[Jodie]

you need to show that the numbers are relevant to paleontology, and that not using them causes problems.

Relevant as in accurate, not using them might cause problems depending on what you are using paleontology for....to understand the past in order to anticipate the future. That might be a big problem for humans if we get it wrong.

 

[Jodie]

The Neanderthals didn't have much genetic variability and lived in small groups at any given time, one would think genetic abnormalities would show up in what is found due to inbreeding as the population dwindled, but that isn't the case.

That's kind of strange unless of course they committed infanticide, then it might not be likely that you would find any fossilized bones since they didn't bury their dead deeply in the ground.