SIMMAP | model parameters

Overview: Molecular Substitution Model Parameters

SIMMAP was written primarily to perform a full hierarchical Bayesian character analysis. Because SIMMAP implements an indirect, or post-tree, character analysis this requires samples of the phylogeny, branch lengths, and subsititution model parameters from the posterior from an external program (e.g., MrBayes). This help document explains how to input, manipulate and use a molecular model parameter file. A detailed discussion of configuring the substitution model can be found here.

Parameter File

SIMMAP reads a column delimited paramter input file. This file is white space delimited (space, tab, etc.) list of parameters in text format. The first line must be a header row (also, white spaced delimited). The header is not case sensitive; pi(A) is equivalant to PI(A). Every subsequent row should contain a value for each header value. The file can contain values for multiple models but SIMMAP will only read values from a single model partition. A comment block is allowed as long as it is contained within an opening square bracket and a closing square bracket (e.g., [comment here]) and occupies the first line only.

Before opening a parameter file SIMMAP needs to be told how to recognize which columns represent different parameters. This is accomplished by configuring the parameter headers (column header string). This should be done before reading the file.

A parameter file can be opened by selecting Open Parameters... in the File menu. The parameter file can be viewed by selecting Show Parameter View in the Window item in the main menu. This opens the following window which shows a small portion of a parameter file from MrBayes.

This is a listing of the contents of the file as referenced by the parameter headers.

The default header values in SIMMAP are those from MrBayes v3.0 (shown below). If these are not the desired headers each parameter can be assigned a unique string by selecting Parameters Headers... in the Models menu. For some additional details on the parameter file configuration see the help documents on input files under the section describing parameter headers. The default header values used are:

Parameter	Header
Freq A	pi(A)
Freq C	pi(C)
Freq G	pi(G)
Freq T	pi(T)
Kappa	kappa
Rate A to C	r(A<->C)
Rate A to G	r(A<->G)
Rate A to T	r(A<->T)
Rate C to G	r(C<->G)
Rate C to T	r(C<->T)
Rate G to T	r(G<->T)
Alpha	alpha
P-Inv Sites	p-inv
Site Specific 1	m{1}
Site Specific 2	m{2}
Site Specific 3	m{3}

Linking Parameters To Tree Samples

Once the parameters (columns) are configured to be identifiable as particular parameters in the model the next step is to link the paremeter samples in the file (rows) to the trees in the tree file. The default in SIMMAP is to set the configuration as a full hierarchical Bayesian design. If the number of parameter samples is NOT equal to the number of trees SIMMAP sets the default to randomly sampling parameter samples for each tree evaluated. This results in configuring the analysis to be as close to a full hierarchical analysis as possible. When this approach is used the autocorrelation between the topology/branch lengths is broken.

While SIMMAP defaults to a full hierarchical analysis, many different kinds of configuration are possible. The options for linking parameters to trees is found in the Parameters tab in the Simulate Histories window (shown to the right).

A couple aspects should be noted first before discussing each of these options: (1) parameter values that are excluded, as described above, only effect options that are starred (*); and (2) these complexity of the design will depend on which trees have been configured to be used in the Trees tab of the Simulate Histories window. For the sake of brevity it will be assumed in this document that Use all trees in file has been selected.

The following options exist for selecting which parameters to use:

Use all values in parameter file *
- All values in the file, except those that have been un-selected, are available for linking directly to the corresponding tree sample or individually to each tree sample. If a parameter value is un-selected then the corresponding tree is skipped if selecting to link each parameter to its corresponding tree. This option is only available if the number of parameters (rows) equals the number of trees (rows). This option is coupled with the Link parameter samples with tree samples * (see Diagram 1 below) and Evaluate parameter values for all trees (see Diagram 2 below) options at the bottom of the tab.
Use randomly chosen parameter values from file *
- This option will make random draws from the parameters except those that have been un-selected. When this option is not set in conjunction with Evaluate parameter for all trees it links to the corresponding tree sample in the tree file (see Diagram 3 below) or the particular tree selected. Otherwise, if Evaluate parameter for all trees is selected (see Diagram 4 below) then it evaluates each random draw against all of the available trees (or as determined in the Trees tab of the Simulate Histories window).
Use values from row
- This option selects a particular row in the parameter file to use regardless of whether it is selected or unselected in the Parameter View window. The row values are evaluated over the trees selected in the Trees tab of the Simulate Histories window.
Use mean values of parameters in file
- The parameter values are summarized as the mean. The user can select to exlcude the values that are unselected in the Parameter View window by selecting the Use only selected rows when calculating. The mean value is evaluated against all selected trees.
Use median values of parameters in file
- The parameter values are summarized as the median. The user can select to exlcude the values that are unselected in the Parameter View window by selecting the Use only selected rows when calculating. The median value is evaluated against all selected trees.

The flow diagrams below provide a graphic depiction of the ways of linking parameters and trees in a character analysis.

Diagram 1: Link Parameter Sample With Tree Samples

Diagram 2: Evaluate Parameter Values For All Trees

Diagram 3: Evaluate Random Parameter/Tree Values

Diagram 4: Evaluate Random Parameter Values Against All Trees


	Help Documentation Current document


	Page Last Updated: 6 August 2008