int main(int argc, char *argv[])
{
{
// test toStringWithComma
int x[] = {INT_MIN, -1234567890, -123456, -12345, -1000, -999, -1,
0, 1, 999, 1000, 12345, 123456, 1234567890, INT_MAX};
int n = sizeof(x)/ sizeof(x[0]);
for (int i = 0; i < n; ++i) {
printf("%-15d : %s\n", x[i], toStringWithComma(x[i]).c_str());
}
}
{
// test toString
std::vector<double> a;
a.push_back(1.1);
a.push_back(2.0);
std::string s = floatToString(a);
printf("%s\n", s.c_str());
}
return 0;
}
int ModelManager::create(const std::string& modelType,
const ModelParser& parser) {
const size_t previousModelNumber = model.size();
std::string modelName = parser.getName();
int nPerm = 10000;
double alpha = 0.05;
int windowSize = 1000000;
if (modelType == "single") {
if (modelName == "wald") {
model.push_back(new SingleVariantWaldTest);
} else if (modelName == "score") {
model.push_back(new SingleVariantScoreTest);
} else if (modelName == "exact") {
model.push_back(new SingleVariantFisherExactTest);
} else if (modelName == "famscore") {
model.push_back(new SingleVariantFamilyScore);
} else if (modelName == "famlrt") {
model.push_back(new SingleVariantFamilyLRT);
} else if (modelName == "famgrammargamma") {
model.push_back(new SingleVariantFamilyGrammarGamma);
} else if (modelName == "firth") {
model.push_back(new SingleVariantFirthTest);
} else {
logger->error("Unknown model name: %s .", modelName.c_str());
abort();
}
} else if (modelType == "burden") {
if (modelName == "cmc") {
model.push_back(new CMCTest);
} else if (modelName == "zeggini") {
model.push_back(new ZegginiTest);
} else if (modelName == "mb") {
parser.assign("nPerm", &nPerm, 10000).assign("alpha", &alpha, 0.05);
model.push_back(new MadsonBrowningTest(nPerm, alpha));
logger->info(
"MadsonBrowning test significance will be evaluated using %d "
"permutations",
nPerm);
} else if (modelName == "exactcmc") {
model.push_back(new CMCFisherExactTest);
} else if (modelName == "fp") {
model.push_back(new FpTest);
} else if (modelName == "rarecover") {
parser.assign("nPerm", &nPerm, 10000).assign("alpha", &alpha, 0.05);
model.push_back(new RareCoverTest(nPerm, alpha));
logger->info(
"Rare cover test significance will be evaluated using %d "
"permutations",
nPerm);
} else if (modelName == "cmat") {
parser.assign("nPerm", &nPerm, 10000).assign("alpha", &alpha, 0.05);
model.push_back(new CMATTest(nPerm, alpha));
logger->info(
"cmat test significance will be evaluated using %d permutations",
nPerm);
} else if (modelName == "cmcwald") {
model.push_back(new CMCWaldTest);
} else if (modelName == "zegginiwald") {
model.push_back(new ZegginiWaldTest);
} else if (modelName == "famcmc") {
model.push_back(new FamCMC);
} else if (modelName == "famzeggini") {
model.push_back(new FamZeggini);
} else if (modelName == "famfp") {
model.push_back(new FamFp);
} else {
logger->error("Unknown model name: [ %s ].", modelName.c_str());
abort();
}
} else if (modelType == "vt") {
if (modelName == "cmc") {
model.push_back(new VTCMC);
} else if (modelName == "price") {
parser.assign("nPerm", &nPerm, 10000).assign("alpha", &alpha, 0.05);
model.push_back(new VariableThresholdPrice(nPerm, alpha));
logger->info(
"Price's VT test significance will be evaluated using %d "
"permutations",
nPerm);
} else if (modelName == "zeggini") {
// TODO
logger->error("Not yet implemented.");
} else if (modelName == "mb") {
logger->error("Not yet implemented.");
} else if (modelName == "analyticvt") {
model.push_back(new AnalyticVT(AnalyticVT::UNRELATED));
} else if (modelName == "famanalyticvt") {
model.push_back(new AnalyticVT(AnalyticVT::RELATED));
} else if (modelName == "skat") {
logger->error("Not yet implemented.");
} else {
logger->error("Unknown model name: %s .", modelName.c_str());
abort();
}
} else if (modelType == "kernel") {
if (modelName == "skat") {
double beta1, beta2;
parser.assign("nPerm", &nPerm, 10000)
.assign("alpha", &alpha, 0.05)
.assign("beta1", &beta1, 1.0)
.assign("beta2", &beta2, 25.0);
model.push_back(new SkatTest(nPerm, alpha, beta1, beta2));
logger->info(
"SKAT test significance will be evaluated using %d permutations at "
"alpha = %g weight = Beta[beta1 = %.2f, beta2 = %.2f]",
nPerm, alpha, beta1, beta2);
} else if (modelName == "kbac") {
parser.assign("nPerm", &nPerm, 10000).assign("alpha", &alpha, 0.05);
model.push_back(new KBACTest(nPerm, alpha));
logger->info(
"KBAC test significance will be evaluated using %d permutations",
nPerm);
} else if (modelName == "famskat") {
double beta1, beta2;
parser.assign("beta1", &beta1, 1.0).assign("beta2", &beta2, 25.0);
model.push_back(new FamSkatTest(beta1, beta2));
logger->info(
"SKAT test significance will be evaluated using weight = Beta(beta1 "
"= %.2f, beta2 = %.2f)",
beta1, beta2);
} else {
logger->error("Unknown model name: %s .", modelName.c_str());
abort();
};
} else if (modelType == "meta") {
if (modelName == "score") {
model.push_back(new MetaScoreTest());
} else if (modelName == "dominant") {
model.push_back(new MetaDominantTest());
parser.assign("windowSize", &windowSize, 1000000);
logger->info(
"Meta analysis uses window size %s to produce covariance statistics "
"under dominant model",
toStringWithComma(windowSize).c_str());
model.push_back(new MetaDominantCovTest(windowSize));
} else if (modelName == "recessive") {
model.push_back(new MetaRecessiveTest());
parser.assign("windowSize", &windowSize, 1000000);
logger->info(
"Meta analysis uses window size %s to produce covariance statistics "
"under recessive model",
toStringWithComma(windowSize).c_str());
model.push_back(new MetaRecessiveCovTest(windowSize));
} else if (modelName == "cov") {
parser.assign("windowSize", &windowSize, 1000000);
logger->info(
"Meta analysis uses window size %s to produce covariance statistics "
"under additive model",
toStringWithComma(windowSize).c_str());
model.push_back(new MetaCovTest(windowSize));
}
#if 0
else if (modelName == "skew") {
int windowSize;
parser.assign("windowSize", &windowSize, 1000000);
logger->info("Meta analysis uses window size %d to produce skewnewss statistics", windowSize);
model.push_back( new MetaSkewTest(windowSize) );
} else if (modelName == "kurt") {
int windowSize;
parser.assign("windowSize", &windowSize, 1000000);
logger->info("Meta analysis uses window size %d to produce kurtosis statistics", windowSize);
model.push_back( new MetaKurtTest(windowSize) );
}
#endif
else {
logger->error("Unknown model name: %s .", modelName.c_str());
abort();
}
} else if (modelType == "outputRaw") {
if (modelName == "dump") {
model.push_back(new DumpModel(prefix.c_str()));
} else {
logger->error("Unknown model name: %s .", modelName.c_str());
abort();
}
} else {
logger->error("Unrecognized model type [ %s ]", modelType.c_str());
return -1;
}
// set parameter and output prefix for newly created models
for (size_t i = previousModelNumber; i < model.size(); ++i) {
model[i]->setParameter(parser);
model[i]->setPrefix(prefix);
if (this->binaryOutcome) {
model[i]->setBinaryOutcome();
} else {
model[i]->setQuantitativeOutcome();
}
}
// create output files
for (size_t i = previousModelNumber; i < model.size(); ++i) {
std::string s = this->prefix;
s += ".";
s += model[i]->getModelName();
if (model[i]->needToIndexResult()) {
s += ".assoc.gz";
fOuts.push_back(new FileWriter(s.c_str(), BGZIP));
fileToIndex.push_back(s);
} else {
s += ".assoc";
fOuts.push_back(new FileWriter(s.c_str()));
}
}
assert(fOuts.size() == model.size());
return 0;
}
