使用org.apache.math3的NormalDistribution类时遭遇java.lang.NoClassDefFoundError

错误重现 #

报错语句：

org.apache.commons.math3.distribution.NormalDistribution dis = new org.apache.commons.math3.distribution.NormalDistribution(0,1);

触发java.lang.NoClassDefFoundError，完整测试代码：

package com.src;
import org.apache.commons.math4.core.jdkmath.*;
import org.apache.commons.math4.legacy.distribution.*;
import org.apache.commons.math4.legacy.distribution.fitting.*;
public class Ztest {
  private boolean srs;
  private boolean independent;
  private boolean largeCounts;
  private double significanceLevel;
  private double testStatistic;
  private double pValue;
  private double nullProportion;
  private CategoricalDataSet data;
  public Ztest(boolean random, boolean percent, boolean big, double alpha, CategoricalDataSet d, double Ho, String direction) {
      srs = random;
      independent = percent;
      largeCounts = big;
      significanceLevel = alpha;
      nullProportion = Ho;
      data = d;
      testStatistic = (data.getProportion() - nullProportion)/data.standardError();
      org.apache.commons.math3.distribution.NormalDistribution dis = new org.apache.commons.math3.distribution.NormalDistribution(0,1);
      if(direction.equals("less than")) {
          pValue = dis.cumulativeProbability(testStatistic);
      }
      else if(direction.equals("greater than")) {
          pValue = dis.cumulativeProbability( -1.0 * testStatistic);
      }
      else {
          if(testStatistic > nullProportion) {
              pValue = 2.0 * dis.cumulativeProbability( -1.0 * testStatistic);
          }
          else {
              pValue = 2.0 * dis.cumulativeProbability(testStatistic);
          }
      }
  }
  public String summary() {
      String message;
      if(pValue < significanceLevel) {
          message = "Since the Pvalue is less than the signficance level, we have sufficient evidence to reject the null hypothesis (H0).";
      }
      else {
          message = "Since the Pvalue is greater than the significance level, we have insufficient evidence to reject the null hypthesis.";
      }
      return message;
  }
}

错误原因 #

1. 依赖版本不匹配：代码中导入的是Apache Commons Math4的包（org.apache.commons.math4.*），但尝试实例化的是Math3的NormalDistribution类，且项目未正确引入Math3的依赖，导致JVM无法找到该类的字节码文件。
2. 包路径混淆：Math4将旧版兼容API放在legacy子包下，其正态分布类路径为org.apache.commons.math4.legacy.distribution.NormalDistribution，无需跨版本使用Math3的类。

解决方法 #

方法一：统一使用Math4 API（推荐） #

1. 修改实例化代码，直接使用已导入的Math4类：

// 因为已经导入了org.apache.commons.math4.legacy.distribution.*，可以直接简写
NormalDistribution dis = new NormalDistribution(0, 1);

2. 确保项目只引入Math4依赖，移除多余的Math3依赖（若存在）。以Maven为例，添加依赖：

<dependency>
    <groupId>org.apache.commons</groupId>
    <artifactId>commons-math4</artifactId>
    <version>4.0</version> <!-- 可替换为最新稳定版 -->
</dependency>

方法二：强制使用Math3 API #

1. 引入Math3的依赖，以Maven为例：

<dependency>
    <groupId>org.apache.commons</groupId>
    <artifactId>commons-math3</artifactId>
    <version>3.6.1</version> <!-- 最新稳定版 -->
</dependency>

2. 移除不必要的Math4依赖，避免类路径冲突；若需同时使用两个版本，需明确区分包路径，避免混淆。

额外代码优化提示 #

双尾检验的逻辑存在错误：当前代码将testStatistic与nullProportion比较，正确逻辑应与0比较，修正后：

else {
    if(testStatistic > 0) {
        pValue = 2.0 * dis.cumulativeProbability(-1.0 * testStatistic);
    } else {
        pValue = 2.0 * dis.cumulativeProbability(testStatistic);
    }
}

内容的提问来源于stack exchange，提问作者Arnav