public static double kruskalWallisPermutationTest(string[] columnNames, IEnumerable[] ie_cofounders, List <string> cof_typeof)
{
//block (intercept) REngine from printing to the Console
//we are just redirecting the output of it to some StringWriter
var stdOut = Console.Out;
Console.SetOut(new StringWriter());
//create the cofounder data frame
regrCovars rCovar = covarsDf(ie_cofounders, cof_typeof);
if (cof_typeof.Count > 0)
{
DataFrame cofounders = rEngineInstance.engine.CreateDataFrame(rCovar.IEcofounders, columnNames: rCovar.cofounders.ToArray());
rEngineInstance.engine.SetSymbol("cofounders", cofounders);
}
//create the data frame
DataFrame df = rEngineInstance.engine.CreateDataFrame(dataFrameValues, columnNames: columnNames);
rEngineInstance.engine.SetSymbol("df", df);
//Approximative (Monte Carlo) multivariate Kruskal-Wallis test
//For two samples (phenotypes), the Kruskal-Wallis test is equivalent to the W-M-W (Wilcoxon-Mann-Whitney) test
rEngineInstance.engine.Evaluate(string.Format(@"kwres <- independence_test(df[,'{0}']{2} ~ factor(df[,'{1}']), teststat = ""quadratic"",
distribution = approximate(B = {3}), ytrafo = function(data) trafo(data, numeric_trafo = rank_trafo))",
columnNames[1],
columnNames[0],
rCovar.cof_string,
publicVariables.numberOfPermutations.ToString()));
rEngineInstance.engine.Evaluate("rm(df)");
//Re-enable Console printings
Console.SetOut(stdOut);
return(Math.Round(rEngineInstance.engine.Evaluate(@"kwres@distribution@pvalue(kwres@statistic@teststatistic)[1]").AsNumeric().First(), 5));
}
public static msMetabolite.stats.regressValues linearRegressionTest(string[] columnNames, string _typeof, IEnumerable[] ie_cofounders, List <string> cof_typeof)
{
//block (intercept) REngine from printing to the Console
//we are just redirecting the output of it to some StringWriter
var stdOut = Console.Out;
Console.SetOut(new StringWriter());
//create the cofounder data frame
regrCovars rCovar = covarsDf(ie_cofounders, cof_typeof);
if (cof_typeof.Count > 0)
{
DataFrame cofounders = rEngineInstance.engine.CreateDataFrame(rCovar.IEcofounders, columnNames: rCovar.cofounders.ToArray());
rEngineInstance.engine.SetSymbol("cofounders", cofounders);
}
//create the data frame
DataFrame df = rEngineInstance.engine.CreateDataFrame(dataFrameValues, columnNames: columnNames);
rEngineInstance.engine.SetSymbol("df", df);
//Exact permutation test intends to run a complete permutation test which is extremely time consuming
//When there are more that 10 observations it will switch from Exact to Prob which approximates the permutation distribution by randomly exchanging pairs of Y elements
//in case we want to use Exact the we should change the limit of Exact from 10 to something greater by using maxExact = X (where X>10)
//nCycle performs a complete random permutation, instead of pairwise exchanges, every nCycle cycles
//run permutation test and take the pvalue
rEngineInstance.engine.Evaluate(string.Format(@"myrestmp <- lmPerm::lmp({0}(df[,'{1}']) ~ as.numeric(df[,'{2}']){3}, perm = ""{4}"", seqs = {5}, " +
@"center = {5}, projections = {6}, qr = {6}, maxIter = {7}, nCycle = {8})",
(_typeof == "factor") ? "as.factor" : ((_typeof == "number") ? "as.numeric" : "Regression failed"),
columnNames[0],
columnNames[1],
rCovar.cof_string,
"Prob",
"FALSE",
"TRUE",
publicVariables.numberOfPermutations,
(publicVariables.numberOfPermutations / 1000).ToString()));
rEngineInstance.engine.Evaluate(@"myres <- summary(myrestmp)");
rEngineInstance.engine.Evaluate(@"myresbeta <- unname(lm.beta::coef.lm.beta(lm.beta::lm.beta(myrestmp))[2])");
//rEngineInstance.engine.Evaluate("print(myres)");
rEngineInstance.engine.Evaluate("rm(df, cofounders, myrestmp)");
//Re-enable Console printings
Console.SetOut(stdOut);
//Console.WriteLine("name " + columnNames.First());
//Console.WriteLine("pv " + Math.Round(rEngineInstance.engine.Evaluate("myres$coefficients[2, 3]").AsNumeric().First(), 5));
//Console.WriteLine("r2 " + Math.Round(rEngineInstance.engine.Evaluate("myres$adj.r.squared").AsNumeric().First(), 5));
//Console.WriteLine("beta " + Math.Round(rEngineInstance.engine.Evaluate("myresbeta").AsNumeric().First(), 5));
return(new msMetabolite.stats.regressValues()
{
clinical_data_name = columnNames.First(),
regrPvalue = Math.Round(rEngineInstance.engine.Evaluate("myres$coefficients[2, 3]").AsNumeric().First(), 5),
regrAdjRsquare = Math.Round(rEngineInstance.engine.Evaluate("myres$adj.r.squared").AsNumeric().First(), 5),
regrBeta = Math.Round(rEngineInstance.engine.Evaluate("myresbeta").AsNumeric().First(), 5)
});
}
