/**
* Write the marginalLikelihoodEstimator, pathSamplingAnalysis and steppingStoneSamplingAnalysis
* blocks.
*
* @param writer XMLWriter
*/
public void writeMLE(XMLWriter writer, MarginalLikelihoodEstimationOptions options) {
if (options.performMLE) {
writer.writeComment("Define marginal likelihood estimator (PS/SS) settings");
List<Attribute> attributes = new ArrayList<Attribute>();
// attributes.add(new Attribute.Default<String>(XMLParser.ID, "mcmc"));
attributes.add(
new Attribute.Default<Integer>(
MarginalLikelihoodEstimator.CHAIN_LENGTH, options.mleChainLength));
attributes.add(
new Attribute.Default<Integer>(
MarginalLikelihoodEstimator.PATH_STEPS, options.pathSteps));
attributes.add(
new Attribute.Default<String>(
MarginalLikelihoodEstimator.PATH_SCHEME, options.pathScheme));
if (!options.pathScheme.equals(MarginalLikelihoodEstimator.LINEAR)) {
attributes.add(
new Attribute.Default<Double>(
MarginalLikelihoodEstimator.ALPHA, options.schemeParameter));
}
writer.writeOpenTag(MarginalLikelihoodEstimator.MARGINAL_LIKELIHOOD_ESTIMATOR, attributes);
writer.writeOpenTag("samplers");
writer.writeIDref("mcmc", "mcmc");
writer.writeCloseTag("samplers");
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>(XMLParser.ID, "pathLikelihood"));
writer.writeOpenTag(PathLikelihood.PATH_LIKELIHOOD, attributes);
writer.writeOpenTag(PathLikelihood.SOURCE);
writer.writeIDref(CompoundLikelihoodParser.POSTERIOR, CompoundLikelihoodParser.POSTERIOR);
writer.writeCloseTag(PathLikelihood.SOURCE);
writer.writeOpenTag(PathLikelihood.DESTINATION);
writer.writeIDref(CompoundLikelihoodParser.PRIOR, CompoundLikelihoodParser.PRIOR);
writer.writeCloseTag(PathLikelihood.DESTINATION);
writer.writeCloseTag(PathLikelihood.PATH_LIKELIHOOD);
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>(XMLParser.ID, "MLELog"));
attributes.add(new Attribute.Default<Integer>("logEvery", options.mleLogEvery));
attributes.add(new Attribute.Default<String>("fileName", options.mleFileName));
writer.writeOpenTag("log", attributes);
writer.writeIDref("pathLikelihood", "pathLikelihood");
writer.writeCloseTag("log");
writer.writeCloseTag(MarginalLikelihoodEstimator.MARGINAL_LIKELIHOOD_ESTIMATOR);
writer.writeComment("Path sampling estimator from collected samples");
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>("fileName", options.mleFileName));
writer.writeOpenTag(PathSamplingAnalysis.PATH_SAMPLING_ANALYSIS, attributes);
writer.writeTag(
"likelihoodColumn", new Attribute.Default<String>("name", "pathLikelihood.delta"), true);
writer.writeTag(
"thetaColumn", new Attribute.Default<String>("name", "pathLikelihood.theta"), true);
writer.writeCloseTag(PathSamplingAnalysis.PATH_SAMPLING_ANALYSIS);
writer.writeComment("Stepping-stone sampling estimator from collected samples");
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>("fileName", options.mleFileName));
writer.writeOpenTag(
SteppingStoneSamplingAnalysis.STEPPING_STONE_SAMPLING_ANALYSIS, attributes);
writer.writeTag(
"likelihoodColumn", new Attribute.Default<String>("name", "pathLikelihood.delta"), true);
writer.writeTag(
"thetaColumn", new Attribute.Default<String>("name", "pathLikelihood.theta"), true);
writer.writeCloseTag(SteppingStoneSamplingAnalysis.STEPPING_STONE_SAMPLING_ANALYSIS);
} else if (options.performMLEGSS) {
// First define necessary components for the tree working prior
if (options.choiceTreeWorkingPrior.equals("Product of exponential distributions")) {
// more general product of exponentials needs to be constructed
if (DEBUG) {
System.err.println("productOfExponentials selected: " + options.choiceTreeWorkingPrior);
}
List<Attribute> attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>(XMLParser.ID, "exponentials"));
attributes.add(new Attribute.Default<String>("fileName", beautiOptions.logFileName));
attributes.add(
new Attribute.Default<String>("burnin", "" + beautiOptions.chainLength * 0.10));
attributes.add(
new Attribute.Default<String>("parameterColumn", "coalescentEventsStatistic"));
attributes.add(
new Attribute.Default<String>(
"dimension", "" + (beautiOptions.taxonList.getTaxonCount() - 1)));
writer.writeOpenTag(
TreeWorkingPriorParsers.PRODUCT_OF_EXPONENTIALS_POSTERIOR_MEANS_LOESS, attributes);
writer.writeTag(
TreeModel.TREE_MODEL,
new Attribute.Default<String>(XMLParser.ID, TreeModel.TREE_MODEL),
true);
writer.writeCloseTag(TreeWorkingPriorParsers.PRODUCT_OF_EXPONENTIALS_POSTERIOR_MEANS_LOESS);
} else {
// matching coalescent model has to be constructed
// getting the coalescent model
if (DEBUG) {
System.err.println(
"matching coalescent model selected: " + options.choiceTreeWorkingPrior);
System.err.println(beautiOptions.getPartitionTreePriors().get(0).getNodeHeightPrior());
}
/*for (PartitionTreePrior prior : options.getPartitionTreePriors()) {
treePriorGenerator.writeTreePriorModel(prior, writer);
writer.writeText("");
}*/
// TODO: extend for more than 1 coalescent model?
TreePriorType nodeHeightPrior =
beautiOptions.getPartitionTreePriors().get(0).getNodeHeightPrior();
switch (nodeHeightPrior) {
case CONSTANT:
writer.writeComment("A working prior for the constant population size model.");
writer.writeOpenTag(
ConstantPopulationModelParser.CONSTANT_POPULATION_MODEL,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "constantReference"),
new Attribute.Default<String>(
"units", Units.Utils.getDefaultUnitName(beautiOptions.units))
});
writer.writeOpenTag(ConstantPopulationModelParser.POPULATION_SIZE);
writeParameter(
"constantReference.popSize",
"constant.popSize",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ConstantPopulationModelParser.POPULATION_SIZE);
writer.writeCloseTag(ConstantPopulationModelParser.CONSTANT_POPULATION_MODEL);
writer.writeComment("A working prior for the coalescent.");
writer.writeOpenTag(
CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "coalescentReference")
});
writer.writeOpenTag(CoalescentLikelihoodParser.MODEL);
writer.writeIDref(
ConstantPopulationModelParser.CONSTANT_POPULATION_MODEL,
beautiOptions.getPartitionTreePriors().get(0).getPrefix() + "constantReference");
writer.writeCloseTag(CoalescentLikelihoodParser.MODEL);
writer.writeOpenTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeIDref(TreeModel.TREE_MODEL, modelPrefix + TreeModel.TREE_MODEL);
writer.writeCloseTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeCloseTag(CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD);
break;
case EXPONENTIAL:
writer.writeComment("A working prior for the exponential growth model.");
writer.writeOpenTag(
ExponentialGrowthModelParser.EXPONENTIAL_GROWTH_MODEL,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "exponentialReference"),
new Attribute.Default<String>(
"units", Units.Utils.getDefaultUnitName(beautiOptions.units))
});
writer.writeOpenTag(ExponentialGrowthModelParser.POPULATION_SIZE);
writeParameter(
"exponentialReference.popSize",
"exponential.popSize",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ExponentialGrowthModelParser.POPULATION_SIZE);
writer.writeOpenTag(ExponentialGrowthModelParser.GROWTH_RATE);
writeParameter(
"exponentialReference.growthRate",
"exponential.growthRate",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ExponentialGrowthModelParser.GROWTH_RATE);
writer.writeCloseTag(ExponentialGrowthModelParser.EXPONENTIAL_GROWTH_MODEL);
writer.writeComment("A working prior for the coalescent.");
writer.writeOpenTag(
CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "coalescentReference")
});
writer.writeOpenTag(CoalescentLikelihoodParser.MODEL);
writer.writeIDref(
ExponentialGrowthModelParser.EXPONENTIAL_GROWTH_MODEL,
beautiOptions.getPartitionTreePriors().get(0).getPrefix() + "constantReference");
writer.writeCloseTag(CoalescentLikelihoodParser.MODEL);
writer.writeOpenTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeIDref(TreeModel.TREE_MODEL, modelPrefix + TreeModel.TREE_MODEL);
writer.writeCloseTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeCloseTag(CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD);
break;
case LOGISTIC:
writer.writeComment("A working prior for the logistic growth model.");
writer.writeOpenTag(
LogisticGrowthModelParser.LOGISTIC_GROWTH_MODEL,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "logisticReference"),
new Attribute.Default<String>(
"units", Units.Utils.getDefaultUnitName(beautiOptions.units))
});
writer.writeOpenTag(LogisticGrowthModelParser.POPULATION_SIZE);
writeParameter(
"logisticReference.popSize",
"logistic.popSize",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(LogisticGrowthModelParser.POPULATION_SIZE);
writer.writeOpenTag(LogisticGrowthModelParser.GROWTH_RATE);
writeParameter(
"logisticReference.growthRate",
"logistic.growthRate",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(LogisticGrowthModelParser.GROWTH_RATE);
writer.writeOpenTag(LogisticGrowthModelParser.TIME_50);
writeParameter(
"logisticReference.t50",
"logistic.t50",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(LogisticGrowthModelParser.TIME_50);
writer.writeCloseTag(LogisticGrowthModelParser.LOGISTIC_GROWTH_MODEL);
writer.writeComment("A working prior for the coalescent.");
writer.writeOpenTag(
CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "coalescentReference")
});
writer.writeOpenTag(CoalescentLikelihoodParser.MODEL);
writer.writeIDref(
LogisticGrowthModelParser.LOGISTIC_GROWTH_MODEL,
beautiOptions.getPartitionTreePriors().get(0).getPrefix() + "constantReference");
writer.writeCloseTag(CoalescentLikelihoodParser.MODEL);
writer.writeOpenTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeIDref(TreeModel.TREE_MODEL, modelPrefix + TreeModel.TREE_MODEL);
writer.writeCloseTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeCloseTag(CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD);
break;
case EXPANSION:
writer.writeComment("A working prior for the expansion growth model.");
writer.writeOpenTag(
ExpansionModelParser.EXPANSION_MODEL,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "expansionReference"),
new Attribute.Default<String>(
"units", Units.Utils.getDefaultUnitName(beautiOptions.units))
});
writer.writeOpenTag(ExpansionModelParser.POPULATION_SIZE);
writeParameter(
"expansionReference.popSize",
"expansion.popSize",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ExpansionModelParser.POPULATION_SIZE);
writer.writeOpenTag(ExpansionModelParser.GROWTH_RATE);
writeParameter(
"expansionReference.growthRate",
"expansion.growthRate",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ExpansionModelParser.GROWTH_RATE);
writer.writeOpenTag(ExpansionModelParser.ANCESTRAL_POPULATION_PROPORTION);
writeParameter(
"expansionReference.ancestralProportion",
"expansion.ancestralProportion",
beautiOptions.logFileName,
(int) (options.mleChainLength * 0.10),
writer);
writer.writeCloseTag(ExpansionModelParser.ANCESTRAL_POPULATION_PROPORTION);
writer.writeCloseTag(ExpansionModelParser.EXPANSION_MODEL);
writer.writeComment("A working prior for the coalescent.");
writer.writeOpenTag(
CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD,
new Attribute[] {
new Attribute.Default<String>(XMLParser.ID, modelPrefix + "coalescentReference")
});
writer.writeOpenTag(CoalescentLikelihoodParser.MODEL);
writer.writeIDref(
ExpansionModelParser.EXPANSION_MODEL,
beautiOptions.getPartitionTreePriors().get(0).getPrefix() + "constantReference");
writer.writeCloseTag(CoalescentLikelihoodParser.MODEL);
writer.writeOpenTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeIDref(TreeModel.TREE_MODEL, modelPrefix + TreeModel.TREE_MODEL);
writer.writeCloseTag(CoalescentLikelihoodParser.POPULATION_TREE);
writer.writeCloseTag(CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD);
break;
default:
// Do not switch to product of exponentials as the coalescentEventsStatistic has not
// been logged
// TODO: show menu that explains mismatch between prior and working prior?
// TODO: but show it when the MCM option is wrongfully being selected, don't do anything
// here
}
}
writer.writeComment("Define marginal likelihood estimator (GSS) settings");
List<Attribute> attributes = new ArrayList<Attribute>();
attributes.add(
new Attribute.Default<Integer>(
MarginalLikelihoodEstimator.CHAIN_LENGTH, options.mleChainLength));
attributes.add(
new Attribute.Default<Integer>(
MarginalLikelihoodEstimator.PATH_STEPS, options.pathSteps));
attributes.add(
new Attribute.Default<String>(
MarginalLikelihoodEstimator.PATH_SCHEME, options.pathScheme));
if (!options.pathScheme.equals(MarginalLikelihoodEstimator.LINEAR)) {
attributes.add(
new Attribute.Default<Double>(
MarginalLikelihoodEstimator.ALPHA, options.schemeParameter));
}
writer.writeOpenTag(MarginalLikelihoodEstimator.MARGINAL_LIKELIHOOD_ESTIMATOR, attributes);
writer.writeOpenTag("samplers");
writer.writeIDref("mcmc", "mcmc");
writer.writeCloseTag("samplers");
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>(XMLParser.ID, "pathLikelihood"));
writer.writeOpenTag(PathLikelihood.PATH_LIKELIHOOD, attributes);
writer.writeOpenTag(PathLikelihood.SOURCE);
writer.writeIDref(CompoundLikelihoodParser.POSTERIOR, CompoundLikelihoodParser.POSTERIOR);
writer.writeCloseTag(PathLikelihood.SOURCE);
writer.writeOpenTag(PathLikelihood.DESTINATION);
writer.writeOpenTag(CompoundLikelihoodParser.WORKING_PRIOR);
ArrayList<Parameter> parameters = beautiOptions.selectParameters();
for (Parameter param : parameters) {
if (DEBUG) {
System.err.println(param.toString() + " " + param.priorType.toString());
}
// should leave out those parameters set by the coalescent
if (param.priorType != PriorType.NONE_TREE_PRIOR) {
// TODO: frequencies is multidimensional, is that automatically dealt with?
writer.writeOpenTag(
WorkingPriorParsers.NORMAL_REFERENCE_PRIOR,
new Attribute[] {
new Attribute.Default<String>("fileName", beautiOptions.logFileName),
new Attribute.Default<String>("parameterColumn", param.getName()),
new Attribute.Default<String>("burnin", "" + beautiOptions.chainLength * 0.10)
});
writeParameterIdref(writer, param);
writer.writeCloseTag(WorkingPriorParsers.NORMAL_REFERENCE_PRIOR);
}
}
if (options.choiceTreeWorkingPrior.equals("Product of exponential distributions")) {
writer.writeIDref("productOfExponentialsPosteriorMeansLoess", "exponentials");
} else {
writer.writeIDref(CoalescentLikelihoodParser.COALESCENT_LIKELIHOOD, "coalescentReference");
}
writer.writeCloseTag(CompoundLikelihoodParser.WORKING_PRIOR);
writer.writeCloseTag(PathLikelihood.DESTINATION);
writer.writeCloseTag(PathLikelihood.PATH_LIKELIHOOD);
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>(XMLParser.ID, "MLELog"));
attributes.add(new Attribute.Default<Integer>("logEvery", options.mleLogEvery));
attributes.add(new Attribute.Default<String>("fileName", options.mleFileName));
writer.writeOpenTag("log", attributes);
writer.writeIDref("pathLikelihood", "pathLikelihood");
writer.writeCloseTag("log");
writer.writeCloseTag(MarginalLikelihoodEstimator.MARGINAL_LIKELIHOOD_ESTIMATOR);
writer.writeComment("Generalized stepping-stone sampling estimator from collected samples");
attributes = new ArrayList<Attribute>();
attributes.add(new Attribute.Default<String>("fileName", options.mleFileName));
writer.writeOpenTag(
GeneralizedSteppingStoneSamplingAnalysis.GENERALIZED_STEPPING_STONE_SAMPLING_ANALYSIS,
attributes);
writer.writeTag(
"sourceColumn", new Attribute.Default<String>("name", "pathLikelihood.source"), true);
writer.writeTag(
"destinationColumn",
new Attribute.Default<String>("name", "pathLikelihood.destination"),
true);
writer.writeTag(
"thetaColumn", new Attribute.Default<String>("name", "pathLikelihood.theta"), true);
writer.writeCloseTag(
GeneralizedSteppingStoneSamplingAnalysis.GENERALIZED_STEPPING_STONE_SAMPLING_ANALYSIS);
}
}
