Statistical Analysis and Evolution

Released from Statistical Analysis and Evolution. --AdminI'm strictly an amateur paleontologist but have devised a statistical test for evolutionary theory. I'm not sure if it's possible, but here's the general proposal.1) Clades for the fossil record are composed for all fossil records within a statistically-significant number of regions around the world. These should already exist for most regions.2) Moderm flora and fauna clades are then "merged" into the fossil record clades. Those that cannot be merged are left as separate clades.3) Step 2 is then repeated but for all the other geographical regions in the study.If evolution is true, then the number clades created in step 2 should be statistically more complete than any of those in step 3."A clade is a biological group of species with a common evolutionary ancestor" (Wikipedia, July, 2005). Cladistics is commonly used in paleontology to classify species by using a set of characteristics (primitive and derived) to construct an evolutionary tree. Clades are usually built using the principle of parsimony to construct the most compact tree.Put simply, I am proposing that modern flora and fauna would be more closely related to prehistoric flora and fauna in their own "region" than to those in those in other regions. For example, the large numbers of primates in Africa coupled with the fact that human or human-like fossils are found there might be shown to be statistically good evidence that humans evolved from lower primates. (This is just a crude example, of course). By incorporating modern organisms into existing prehistoric clades by region, then doing this recursively for all sub-regions, it should be possible to show that regions with contemporary organisms and associated fossil record clades are more statistically complete than other regions. Of course, the analysis would have to account for known migrations of species and artificial introduction of species into a region by man and other known factors that would skew the statistics.This would be a huge effort requiring experts in cladistics, paleontology, biology and computer science. As a proposed topic, I'm looking for thoughts on whether this might be worth the effort and if so, how it might provide solid statistical evidence for evolution.This message has been edited by futzman, 07-23-2005 10:04 AMThis message has been edited by futzman, 07-23-2005 01:09 PM

Hi futzman,Are you suggesting that the extant taxa and fossil taxa are merged in a single phylogeny based on fossil evidence for the former and skeletal evidence for the latter? I'm sure these types of phylogenies must exist already (at least one example is here.In fact there are (bayesian) methods of reconstructing phylogenies that would permit you to use fossil evidence only for the extinct species, and a mixture of skeletal and molecular evidence for extant species, therefore taking advantage of the maximum possible amount of data.I would imagine that constructing a single phylogeny for a well known group would be the easiest way to start. I'm not sure I agree with your prediction. Surely it depends on whether speciation occurred after colonization of a new region, or before it. For example if you have a mainland population of many species of rodent, and one single species establishes a new population on an isolated island, then that island population subdivides in situ to produce novel species, then your prediction will be satisfied. All of the extant rodent species on the island will be "more similar" to the fossil island species. But on the other hand, if a large number of species migrate from the mainland to the island, then any novel species produced after the initial migration may be "more similar" to the ancestral mainland stock rather than to the fossil remnants of other novel island species that share the geographic location but have different ancestry.Does that make sense? I often wish I could somehow sketch a picture on the screen in cases like this!Mick

Yes, that's what I'm proposing but the clades would have to be built by arbitrary, possibly recursive, regions. Certainly these possibilities exist. What I'm proposing is that statistically the regional, composite clades would support evolution despite the exceptions. This, of course, may or may not be true. The only way to know would be to run the experiment. Since I'm a software engineer by profession, I can see this requiring a massive amount of computing power and huge amounts of data. However, the results might be worth the effort.

Hi futzman, It seems likely that rapid migration (i.e. the "exceptions") will only be a statistically significant problem in recent clades with densely packed living species and extinct taxa (and even then, may not be a problem at all). You might well end up with a more coherent story at deeper levels in the tree (although it's worth bearing in mind that amongst mammals, a clade that has been pretty well studied in terms of fossils, we still don't know whether the clade arose in gondwanaland or laurasia).I guess you already know that such studies are routinely carried out for living species (molecular studies that by necessity exclude extinct taxa). An encouraging thing is that the results are often consistent with your prediction (geographic-based clades).Here's a nice example from the shrew Sorex monticolus. Every terminal taxa in the tree is monticolus unless otherwise indicated. You can identify distinct geographic clades: at a deep level, a continental clade and a coastal clade; and at a more shallow level, the continental clade is divided into distinct northern and southern groupings. Also interesting is the fact that many monticolus is paraphyletic - the shrews along the coast and in parts of the interior appear to be a kind of biological continuum rather than distinct species. I guess you simply want to do similar studies, at a larger scale, including extinct as well as extant taxa.You're right, the computational problems will be immense. Systematists are facing huge problems just with constructing a phylogenetic tree that contains all of the living mammals, never mind all of the extinct ones as well!Mick

Just as a sidenote, on the question of merging clades.You may be interested in some of the literature on building supertrees. There is a bibliography here. Perhaps you already know about this stuff, otherwise i could fill you in. Certainly if you are a computery type of person you will be interested in the supertree problem. And it may be a way of doing what you want to do.mick

No, I didn't know about these references, so thanks Mick. I'm strictly an amateur paleontologist and came up with this idea one hot afternoon staring at some fossil plant material I had just collected that seemed to match the shape of local willow tree leaves. I am, however, interested in reading any material relating to this problem.

Futzman,
Hate to burst your bubble but this type of analysis has been around for quite a while and yes it can be done to compare fossils and extant species. It goes by the name of numerical taxonomy or phenetics. Google either of those terms and you will get way more than you want to know.

I don't think you read my proposal closely enough. Could you explain how phenetics fits into my proposal?

Sure, a phenetic analysis is done for the fossils in a given region and for the extant species in a given region. The genetic distance is calculated for some set of characters that can be measured in both fossils and extant species (skeletal characters for example). This is done for several regions. Genetic distances should be less within a region than between regions for a given set of characters. (Phenetic analysis produces clades based on the genetic distances)On edit: To answer an earlier point you raised - phenetic analysis requires a great deal of computer power. In the good old days a super computer with parallel processing capability was needed - of course it is no big deal today.This message has been edited by deerbreh, 07-25-2005 04:56 PM

So, if it's been done why is it not used for an argument "for" evolution? Have the results been compiled and published for many regions and flora and fauna? I would think these type of statistics would be veritable definitive proof that macro-level evolutionary processes have taken place.

Taxonomists don't spend a great deal of time worrying about making arguments "for" evolution. They aren't interested in wasting valuable research time convincing a few creationists. I am sure most of this analysis has been done where possible to show relationships between extinct and extant species. However, a lot more can be done to show relationships within, among and between groups of extant species, because the phenetic analysis can be done on DNA characters - a far richer source of hereditary information than the scraps available in fossils. This is how we can be sure that chimpanzees are much more closely related to humans than are other primates for example.

Somehow I knew you were going to say this but I can dream, can't I?

Heh. Here is one link to a study of lizard evolution which uses some of this type of analysis. Warning - very heavy reading!
http://www.serpenti.it/rivista/pubblica4/origin.htm

Hey futzman, interesting idea, and I like how you arranged the question as a test with predictions rather than explanations.Can you elaborate on the regional branching of clades? In particular, the emergence of new lineages with few transitional forms? For example, say the appearance of angiosperms in the Late Jurassic/Early Cretaceous?Also, how would such an analysis measure up against the founder effect and subsequent adaptive radiation? Small populations are near-invisible to the larger resolution of the fossil record. As such, the data may tell a different story than the one proposed.Perhaps... just to feed the idea a bit more, instead of taking a specific regional analysis and adding them up, look at a specific environments and the transitions of flora and fauna within them in time. Look at it through multiple preserved environments to build up your data base, and look for general trends. I will offer the following hypothesis in favor of evolution: There should be high numbers of analogous traits (i.e. dorsal fins on whales and sharks)built from homologous parts known from genetic relatives (the presence of ulnae and radius in bats and mice), as the temporal distance becomes greater. Alternatively, we could postulate a hypothesis in favor of creationism: There should be little variation within set structures for given animals present in the analysis (i.e. all moluscs have a calcium-carbonate shell regardless of what time period). Of course, it is more difficult to support the latter, but perhaps a statistical analysis would shed some light on that, particularly in specific structures known largely by those who study them.I suspect the analysis would support the former evolutionary hypothesis overwhelmingly, but I don't want to throw water on your fire. I suppose you should narrow your hypothesis to specific taxa, if nothing else migration would cancel out a set regional record of fauna and florae (under both creation and evolution models).A tout a l'heurre,
Theus

---

Veri Omni Veritas

Hi Theus.I'm not sure I can answer your questions as your knowledge probably exceeds mine. But I'll try to clarify my hypothesis a bit and you can draw your own conclusions.In a nutshell what I'm proposing is that if I find fossil A in region X and I find an extant species Z in region X that is an apparent descendant (via cladistic analysis) for A in region X, then the clade for A to Z will be more complete than in any other region where A is found. Note that I'm saying "statistically" rather than "always". Of course, there will be many cases where this doesn't hold true, but statistically I'm proposing it will hold true for most cases. Regional recursion may be necessary because the regional areas may not be large enough to create clades that are compete enough for meaningful comparisons.Whether this would hold true for a limited taxa over an isolated region is questionable, of course. My guess is that doing this kind of analysis for marine invertebrates would be daunting (to say the least) but might be possible for vertebrates from say the Triassic to the present (my naivete is probably showing here). The only way I can see my hypothesis would not be true is if dispersion of the species over geographic regions achieved perfect distribution over time, a statistic improbability in my estimation. However, I could of course be all wet on this I suspect your idea about using specific environments rather than geographic regions would probably show even more statistical correlation (in favor of evolution), although I'm guessing the creationists would view this as "cooking the books" and maybe with some merit.FutzThis message has been edited by futzman, 08-09-2005 12:22 AM