Loading ...
Sorry, an error occurred while loading the content.

Phylogenetic analysis results for patristic data

Expand Messages
  • yennifmit
    Hi All, A number of phylogenetic analysis results based on patristic data sets are now available at my Views site:
    Message 1 of 8 , Dec 6, 2013
    • 0 Attachment

      Hi All,


      A number of phylogenetic analysis results based on patristic data sets are now available at my "Views" site:


      http://www.tfinney.net/Views/phylo/nj/Matt-Brooks.png

      http://www.tfinney.net/Views/phylo/nj/Matt-Cosaert.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Matt-Ehrman.png

      http://www.tfinney.net/Views/phylo/nj/Matt-Racine.png

      http://www.tfinney.net/Views/phylo/nj/Mark-Cosaert.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Mark-Mullen.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Luke-Cosaert.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/John-Brooks-C.png

      http://www.tfinney.net/Views/phylo/nj/John-Cosaert.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/John-Cunningham.png

      http://www.tfinney.net/Views/phylo/nj/John-EFH.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Acts-Donker.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Acts-Osburn.png

      http://www.tfinney.net/Views/phylo/nj/Paul-Brooks.png

      http://www.tfinney.net/Views/phylo/nj/Paul-Cunningham.png

      http://www.tfinney.net/Views/phylo/nj/Paul-Donker.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Paul-Osburn.png

      http://www.tfinney.net/Views/phylo/nj/Rom-Donker.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Cor-A-Donker.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Paul-Donker-Cor-B-Titus.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Heb-Donker.15.SMD.png


      These relate to Church Fathers as follows:


      Brooks (NTGF 2, 1991): Gregory of Nyssa

      Cosaert (NTGF 9, 2008): Clement of Alexandria

      Cunningham (PhD diss., 1995): Cyril of Alexandria

      Donker (NTGF 8, 2011): Athanasius of Alexandria

      EFH (NTGF 3, 1992): Origen

      Ehrman (NTGF 1, 1986): Didymus the Blind

      Mullen (NTGF 7, 1997): Cyril of Jerusalem

      Osburn (NTGF 6, 2004): Epiphanius of Salamis

      Racine (NTGF 5, 2004): Basil of Caesarea


      Most of the data sets are from the SBL New Testament in the Greek Fathers series. (EFH = Ehrman, Fee, and Holmes: the data collected by the authors was originally analysed by Bruce Morrill; Jared Anderson later presented the data in his 2008 ThM thesis and that's where I got hold of it.) Richard Mallett deserves many thanks for creating the majority of underlying computer files that were subsequently processed by my R scripts.


      These results were obtained using Naruya Saitou and Masatoshi Nei's Neighbor-Joining method ("Neighbor-Joining Method" _Molecular Biology and Evolution_ 4/4, 1987, 406-425) which seeks to find the minimum-evolution phylogenetic tree, being that tree which most economically accounts for the observed set of distances between objects. While the method "produces a unique final tree under the principle of minimum evolution," it does not always produce the minimum-evolution tree. However, computer simulations show that it "is quite efficient in obtaining the correct tree topology."


      As reported previously, applying NJ analysis to data generated by a manuscript copying simulation produced a tree in which the initial text (which is known in the simulation) was close to the junction of major branches of that artificial tradition. (See http://groups.yahoo.com/neo/groups/textualcriticism/conversations/messages/8040.) That result supports the thesis that the closest witnesses we have to the initial text of the actual tradition are those which stand closest to the junction of the major branches of the extant tradition's minimum-evolution tree. These major branches correspond to the traditional Alexandrian, Western, and Byzantine divisions. There is also another variety, in Mark at least, which corresponds to Streeter's "Eastern type" and includes texts he associated with Syrian Antioch and Caesarea. (See e.g. http://www.tfinney.net/Views/phylo/nj/Mark-UBS4.15.SMD.png.)


      As it happens, the NJ analysis results for patristic data sets listed above show that the corresponding patristic texts tend to branch from the tradition in the vicinity of the junction of major branches.


      The junction of major branches is in the vicinity of texts such as Family 1, 33, 700, 892, and the Syriac versions. See e.g. the following results based on UBS2 or UBS4 apparatus entries for the Gospels:


      http://www.tfinney.net/Views/phylo/nj/Matt-UBS2.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Mark-UBS4.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/Luke-UBS2.15.SMD.png

      http://www.tfinney.net/Views/phylo/nj/John-UBS2.15.SMD.png


      Best,


      Tim Finney


    • Stephen Carlson
      ... Despite the claims of its inventors, Neighbor-Joining isn t really a phylogenetic method. Rather, it is a phenetic method that groups on some measure of
      Message 2 of 8 , Dec 9, 2013
      • 0 Attachment
        On Fri, Dec 6, 2013 at 10:48 AM, <tjf@...> wrote:

        A number of phylogenetic analysis results based on patristic data sets are now available at my "Views" site:

        These results were obtained using Naruya Saitou and Masatoshi Nei's Neighbor-Joining method ("Neighbor-Joining Method" _Molecular Biology and Evolution_ 4/4, 1987, 406-425) which seeks to find the minimum-evolution phylogenetic tree, being that tree which most economically accounts for the observed set of distances between objects. While the method "produces a unique final tree under the principle of minimum evolution," it does not always produce the minimum-evolution tree. However, computer simulations show that it "is quite efficient in obtaining the correct tree topology."

        Despite the claims of its inventors, Neighbor-Joining isn't really a phylogenetic method. Rather, it is a phenetic method that groups on some measure of overall similarity in a manner intended to produce results that might approximate those of a proper phylogentic method. It can only do so, however, if the data are well-behaved, which they really aren't for manuscript data. I don't mind people experimenting with Neighbor-Joining (after all, I was doing so a dozen years ago), but its severe limitations with the messy manuscript data should not be ignored, nor should its results be called "phylogenetic" as if its poor results are in any way representative of proper phylogenetic methods.
         
        Stephen
        --
        Stephen C. Carlson, Ph.D. (Duke)
        Post-Doctoral Fellow, Theology, Uppsala
         
      • Jac Perrin
        Dear Stephen, So what do you prefer in lieu of NJoining? Jac Perrin Sent from my iPhone
        Message 3 of 8 , Dec 10, 2013
        • 0 Attachment
          Dear Stephen,

          So what do you prefer in lieu of NJoining?

          Jac Perrin

          Sent from my iPhone

          On Dec 9, 2013, at 2:27 AM, Stephen Carlson <stemmatic@...> wrote:

           

          On Fri, Dec 6, 2013 at 10:48 AM, <tjf@...> wrote:

          A number of phylogenetic analysis results based on patristic data sets are now available at my "Views" site:

          These results were obtained using Naruya Saitou and Masatoshi Nei's Neighbor-Joining method ("Neighbor-Joining Method" _Molecular Biology and Evolution_ 4/4, 1987, 406-425) which seeks to find the minimum-evolution phylogenetic tree, being that tree which most economically accounts for the observed set of distances between objects. While the method "produces a unique final tree under the principle of minimum evolution," it does not always produce the minimum-evolution tree. However, computer simulations show that it "is quite efficient in obtaining the correct tree topology."

          Despite the claims of its inventors, Neighbor-Joining isn't really a phylogenetic method. Rather, it is a phenetic method that groups on some measure of overall similarity in a manner intended to produce results that might approximate those of a proper phylogentic method. It can only do so, however, if the data are well-behaved, which they really aren't for manuscript data. I don't mind people experimenting with Neighbor-Joining (after all, I was doing so a dozen years ago), but its severe limitations with the messy manuscript data should not be ignored, nor should its results be called "phylogenetic" as if its poor results are in any way representative of proper phylogenetic methods.
           
          Stephen
          --
          Stephen C. Carlson, Ph.D. (Duke)
          Post-Doctoral Fellow, Theology, Uppsala
           

        • Stephen Carlson
          ... For a phylogenetic method, it has to, by definition, take into account derived similarities, so I would prefer a method such as maximum parsimony or
          Message 4 of 8 , Dec 10, 2013
          • 0 Attachment
            On Tue, Dec 10, 2013 at 5:10 PM, Jac Perrin <jperrin@...> wrote:
            So what do you prefer in lieu of NJoining?

            For a phylogenetic method, it has to, by definition, take into account derived similarities, so I would prefer a method such as maximum parsimony or maximum likelihood. These methods, of course, are much more computationally expense and therefore simpler methods such as neighbor joining have their place, as long as the results are not over-interpreted.

            Stephen
            --
            Stephen C. Carlson, Ph.D. (Duke)
            Post-Doctoral Fellow, Theology, Uppsala

          • yennifmit
            Dear Stephen, Thanks for the correction on nomenclature. From now on I will call NJ a phenetic technique. It seems a tad pessimistic to say that NJ is poor or
            Message 5 of 8 , Dec 11, 2013
            • 0 Attachment

              Dear Stephen,


              Thanks for the correction on nomenclature. From now on I will call NJ a phenetic technique.


              It seems a tad pessimistic to say that NJ is poor or not proper. It should also be said that phylogenetic results are not immune from criticism: they are not guaranteed to give the correct tree topology. I suppose it would be prudent to say that the NJ results are provisional -- innocent until proven guilty. It remains true that NJ gave the correct tree topology when applied to the simulation data mentioned in post 8040, placing "texts" with like pedigrees in the same branches. The copying simulation which generated those "texts" includes a mechanism which produces textual mixture, so it seems that NJ works in the presence of mixture.


              Best,


              Tim Finney


            • richardmallett2001
              Does the NJ method produce results that are compatible with your other methods ? (It appears that I can reply to posts via the website; but if I start a new
              Message 6 of 8 , Dec 17, 2013
              • 0 Attachment
                Does the NJ method produce results that are compatible with your other methods ?

                (It appears that I can reply to posts via the website; but if I start a new topic, it appears to work, but the message
                doesn't appear)

                -- 
                --
                Richard Mallett
                Eaton Bray, Dunstable
                South Beds. UK
              • yennifmit
                Hi Richard, There is a broad correspondence between branches seen in the NJ analysis results and branches or groups revealed by other analysis methods. Taking
                Message 7 of 8 , Dec 19, 2013
                • 0 Attachment

                  Hi Richard,


                  There is a broad correspondence between branches seen in the NJ analysis results and branches or groups revealed by other analysis methods. Taking the UBS4 data set for Mark as an example, CMDS, DC, and NJ results have clusters or branches corresponding to Alexandrian, Byzantine, Western, and Streeter's Eastern texts:


                  http://www.tfinney.net/Views/cmds/Matt-UBS4.15.SMD.gif

                  http://www.tfinney.net/Views/dc/Mark-UBS4.15.SMD.png

                  http://www.tfinney.net/Views/phylo/nj/Mark-UBS4.15.SMD.png


                  A five-way PAM partition reveals these four branches and a Vulgate branch too:


                  B: UBS Aleph B L Delta Psi 2427 cop-sa cop-bo

                  Byz: A C f-13 33 157 180 579 597 700 892 1006 1010 1071 1241 1243 1292 1342 1424 1505 Byz E F G H N Sigma Lect syr-p syr-h slav

                  it-ff-2: D it-a it-b it-c it-d it-ff-2 it-i it-k it-r-1

                  f-1: W f-1 28 205 565 syr-s arm geo

                  vg: Theta it-aur it-f it-l it-q vg syr-pal eth Augustine 


                  The witness before the colon at the start of each group is the medoid (i.e. most central member) of that group.


                  Looking at the CMDS, DC, NJ, and PAM results for this data set shows that there is a common theme though a number of witnesses can shift allegiance (e.g. in one branch of an NJ result but not the corresponding group of a PAM partition). Texts which tend to shift represent unusual combinations of readings. I think that might be due to them having mixed texts or being akin to texts from the time before major textual streams developed.


                  Best,


                  Tim Finney

                • richardmallett2001
                  I couldn t reply yesterday. I just wanted to say that that was encouraging that you had so much agreement, and that it would not be surprising if the others
                  Message 8 of 8 , Jan 2, 2014
                  • 0 Attachment
                    I couldn't reply yesterday.  I just wanted to say that that was encouraging that you had so much agreement, and that it would not be surprising if the others were mixed text.


                    -- 
                    --
                    Richard Mallett
                    Eaton Bray, Dunstable
                    South Beds. UK
                  Your message has been successfully submitted and would be delivered to recipients shortly.