Just adding a little clarification to some of the answers. Phenetics does not return any hypotheses of evolutionary relationships, but simply similarities of distance and scale, or geometry, depending on the methods used. Also, categorical data such as colour, or meristic counts, are not morphometric data. They can also be treated phenetically, and the two are sometimes combined in phenetic analyses, but the data are statistically quite different because of different underlying distributions. You have to be very careful statistically when analyzying morphometric data to understand the type of data that you are dealing with, the types of analyses which can be applied to your data legitimately, and the types of hypotheses thus generated. You are free to construct phylogenetic hypotheses based on your morphometric data, but morphometric analysis cannot test those hypotheses. Someone already mentioned gap or discontinuity coding and then application of PAUP, which does result in a phylogenetic analysis. But gap coding itself can be problematic.

If you can encode your data into discrete form you should be able to use Phylip for a phenetic analysis.

If you want to use your morphometric data to infer phylogenetic relationships, I advise you to use TNT. There are some recent implementations that allow for the use of continuous character as input in a phylogenetic analysis.

You should first check in all taxa involved the overall variability and overlapping between them in each of your continuous traits, in order to differentiate the states for each of characters and to assign them. Then you should decide on their nature (ordered, or unordered), i.e., on polarization (transformation series), if you have the means to assess it with remarkable degree of certainty. For delimitation of ranges for decision about character states and ways how to do that I suggest you the old reference of Thorpe, and the software I recommend is PAUP, though I believe that other newer ones would work the job out, as well.

you can first list out all the possible characters and also the character states. Data set can be build under NEXUS editor and analysis can be performed using PAUP.

Read Catalano & Goloboff's paper on using landmark data for cladistic analyses.

Why using phenetic analysis?

I think it is recommendable to use Cladistic analysis. If you decide to do so, consider the use of TNT, a free software available at the Willi Hennig Society's site:

http://www.cladistics.org/tnt.html

A phenetic analysis will not return a phylogeny. To do this you need a cladistic analysis. The dataset can be built with Mesquite, the continuous characters encoded into discrete states with MorphoCode and the analysis performed with PAUP.

There is a related discussion going on in the Taxacom Digest, Vol 80, Issue 14 that might help you out.

Thank you all for the warm responds and comments. I really appreciate them and would consider trying which is best for my analysis.

To eliminate samples size variation among individuals, gerenally speaking, morphometric data need be log-normalized, and transform data can analysised with SPSS software, you can do some hierarchical cluster analysis.

phenetic and phylogenetic are 2 very different approaches but it would be interesting to compare the results of the 2, because in any caes, they are only hypotheses of relationships.

You will find a lot of softwares and algorithms to do them. many free of charge.

For phylogenetics:

http://evolution.genetics.washington.edu/phylip/software.html

For phenetics or morphometrics (or numerical taxonomy):

http://life.bio.sunysb.edu/morph/index.html

Cheers,

Nicolas

Thank you Dr Nicolas. I just realised that they are actually two different approaches. Anyway that's actually what I wanted to do, to compare both of them in term of species relationship. I have done phylogenetics using molecular data, and now I want to compare with the phenetics analysis

Phenetics uses simple evolutionary models to compare the number of morphometric characters chosen for the analysis and how many have matched between the species. In turn, molecular sequence based analysis are robust and many sophisticated models have developed during the recent times so that you can use them to track a little diversity which may not reflect countable morphological characters. I just wanted to say is; if the species chosen are sufficiently diverged to visualize a number of morphological characters then it is worth to consider them and compare with the DNA/protein/any gene based phologenetic tree for confirmation or any further interpretations depends up on your objectives....

Good luck

Morphometric variables include categorical variables and measurements variables(numerical variable), categorical variables contain body colour or number of setae, ect, for example insect elytral color was also used as a categorical variable, with the value “1” indicating black and “0” indicating yellow, Cluster analysis of the measurements and categorical variables was performed by SPSS software.

Phenotype and phylogenetic are tow very different methods but it would be interesting to compare the results of the 2,

For Phenotype: u can use Bergey's manual of determinative bacteriology.

For Phylogenetic analysis: U must have the 16s rRNA and u can go to NCBI and u can get the similarities of the strains nearby ur strain.

Use Bioedit or Mrega programm ( U can download them from Googleor from NCBI website.) to make the phylogenetic tree.

The phylogenetic tree will shows u the closed strain to ur strain.

Regards

Thank you for all suggestions. I have done my phylogenetic analysis using molecular data and now will compare with results using morphometric parameters using cladistic method.

Just adding a little clarification to some of the answers. Phenetics does not return any hypotheses of evolutionary relationships, but simply similarities of distance and scale, or geometry, depending on the methods used. Also, categorical data such as colour, or meristic counts, are not morphometric data. They can also be treated phenetically, and the two are sometimes combined in phenetic analyses, but the data are statistically quite different because of different underlying distributions. You have to be very careful statistically when analyzying morphometric data to understand the type of data that you are dealing with, the types of analyses which can be applied to your data legitimately, and the types of hypotheses thus generated. You are free to construct phylogenetic hypotheses based on your morphometric data, but morphometric analysis cannot test those hypotheses. Someone already mentioned gap or discontinuity coding and then application of PAUP, which does result in a phylogenetic analysis. But gap coding itself can be problematic.

Thank you Peter for the clarifications. Appreciate them very much and will take your points into consideration.

Fink and coauthors proposed many years ago that partial warps from a thin-plate spline geometric morphometric analysis could be used as cladistic characters, but they retracted their position later. You cannot do that, because partial warps depend on the number of landmarks, i.e. they are totally artificial.

Many ways to solve your problem.

The simple one:

Just create a matrix with rankings. For example for a size: 0 for 0cm; 1 for 0 to to 4 cm; 2 for 4 to 10cm etc.... Of course you've got to adapt coding depending on your data (checking distribution patterns).

Then you can use Paup and perform the analysis in phenetic but better to use ordered parsimony (meaning that 0 is higher than 1, 2 etc......). Like this you may use really your quantitative values.

This approach will work...but is not optimal.

The complicate way can be to use Neuron network (SOM for example) which will use vectorial distances. I used these network for genetics but better to check publications of Park or Lek or Brosse or Cereghino who were using it for ecology and morphometry. Then you can build a tree following the Ward algorithm which will give you relationships of distances between your samples (take care...it doesn't mean phylogeny...just distances....that's phenetic).

I'm sure that some canonical analysis will also work.

Cheers from Luc

Thanks Luc!

Hi all,

Many answers have been given to Haslawati's question which spur my interest in posing a related question with the hope of getting reply soonest. Because I have been encountering similar problem of coding continuous morphometric data for phenetic analysis, but my question is which of the coding methods is robust and can be statistically tested? Is it ranging, cut-off, gap coding, gap weighting or step-matrix gap weighting?