基于Welford法的C++标准差计算程序异常：均值与结果排查

Let's break down exactly what's going wrong in your code, plus implement the required Welford's online algorithm correctly—no arrays needed, just a single pass through the data.

1. Input Logic is Broken (Biggest Culprit!) #

Right now, your code is skipping the first score and counting the -1 end marker as a valid grade:

• You read the first score into grade, then immediately read another score at the start of the loop. That first score never gets added to your sum!
• When you type -1 to end input, the code still adds it to sum and sumSq, and increments numtests. That's why your average is way off.

2. Standard Deviation Formula is Wrong #

Your current calculation has two critical math errors:

• Misplaced parentheses: The correct sample variance formula is (sumSq - n*avg²)/(n-1), but your code divides only the n*avg² term by n-1—totally messing up the calculation.
• Missing square root: Standard deviation is the square root of variance, but you never call sqrt() to get the final value!
• No edge case handling: If you enter 0 or 1 scores, you'll hit a divide-by-zero error.

3. You're Not Using Welford's Method #

Your code uses the sum/sum-of-squares approach, but the assignment specifically asks for Welford's online algorithm. This method is better for numerical stability (especially with large datasets) and works in a single pass without storing all data. Here's the core logic:

• Track the current mean and a running variance accumulator (M2) instead of raw sums.
• For each new score, update the mean incrementally, then adjust M2 using the difference between the new score and the old/new mean.

Fixed Code (Welford's Method + Correct Input Handling) #

#include <iostream>
#include <cmath>
using namespace std;

int main()
{
    double grade = 0;
    int numtests = 0;
    double mean = 0.0;
    double M2 = 0.0; // Welford's variance accumulator

    cout << "Enter the test scores. Once finished, enter -1 to end input.\n";
    
    // Only process scores that aren't the end marker (-1)
    while (cin >> grade && grade != -1)
    {
        numtests++;
        double delta = grade - mean;
        mean += delta / numtests;
        double delta2 = grade - mean;
        M2 += delta * delta2;
    }

    cout << "The number of scores: " << numtests << "\n";
    
    // Handle edge cases to avoid crashes/invalid results
    if (numtests == 0)
    {
        cout << "No test scores were entered.\n";
        return 0;
    }

    double average = mean;
    double std_dev;
    if (numtests == 1)
    {
        std_dev = 0.0; // A single score has no meaningful sample standard deviation
    }
    else
    {
        std_dev = sqrt(M2 / (numtests - 1));
    }

    cout << "Average: " << average << "\n";
    cout << "Std. Deviation: " << std_dev << "\n";

    return 0;
}

What Changed? #

• Input Fix: The loop now checks if the input is -1 before processing the score, so the end marker is ignored, and the first score gets counted properly.
• Welford's Implementation: We track mean and M2 iteratively—no sums, no arrays, just a single pass through the input as required.
• Edge Cases: We handle empty input and single-score input to avoid crashes and invalid calculations.

内容的提问来源于stack exchange，提问作者Larry the Guy