Chad Young created MATH-1644:
--------------------------------
Summary: BinomialTest returns p-values > 1
Key: MATH-1644
URL: https://issues.apache.org/jira/browse/MATH-1644
Project: Commons Math
Issue Type: Bug
Affects Versions: 3.6.1
Reporter: Chad Young
The code below produces the output "P-value for 10 trials, 5 successes,
probability 0.5, two-sided alternative is 1.2460937500000002", but p-values
should never be greater than 1. I think what is happening is that the test is
adding both one-sided alternatives, but both include the 5 successes
probability, so it is counted twice. This happens whenever the number of
successes is exactly what is expected based on number of trials and success
probability (numberOfSuccesses = numberOfTrials * probability).
{code:java}
import org.apache.commons.math3.stat.inference.*;
public class Main {
public static void main(String args[]) {
BinomialTest bt = new BinomialTest();
Double pval = bt.binomialTest(10, 5, 0.5,
AlternativeHypothesis.TWO_SIDED);
System.out.println("P-value for 10 trials, 5 successes, probability
0.5, two-sided alternative is " + pval);
}
}
{code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
