Normal Distribution

Understand the normal distribution, its key parameters, and how to generate and analyze it using Python's NumPy library.

What it is

The normal distribution, often called the Gaussian distribution or "bell curve," is a continuous probability distribution that is symmetric around its mean. It describes how data clusters around an average value, with fewer occurrences as you move further away from the center. The shape is defined by two parameters: the mean (μ), which determines the center of the curve, and the standard deviation (σ), which controls the spread or width of the curve.

Key related terms include variance (the square of the standard deviation) and the Central Limit Theorem, which states that the sum of many independent random variables tends toward a normal distribution, regardless of their original distributions.

Why it matters

• Natural Phenomena: Many real-world measurements, such as human heights, IQ scores, and measurement errors, naturally follow a normal distribution.
• Statistical Inference: Most hypothesis tests (like t-tests and ANOVA) assume data is normally distributed or rely on the Central Limit Theorem for large samples.
• Machine Learning: Initializing neural network weights and modeling residuals in regression analysis often assumes normality.
• Risk Management: Financial models use normal distributions to estimate potential losses and volatility.

Syntax or steps

To work with normal distributions in Python, we primarily use the numpy.random.normal function. This function generates random numbers from a normal distribution.

The basic syntax is:

np.random.normal(loc=0.0, scale=1.0, size=None)

• loc: The mean (μ) of the distribution. Default is 0.
• scale: The standard deviation (σ). Must be non-negative. Default is 1.
• size: The number of samples to generate. If None, a single float is returned.

Example

This example generates 1,000 samples from a standard normal distribution (mean=0, std dev=1) and calculates basic statistics to verify the properties.

import numpy as np

# Generate 1000 samples from a normal distribution
# Mean (loc) = 0, Standard Deviation (scale) = 1
samples = np.random.normal(0, 1, 1000)

# Calculate statistics
mean_val = np.mean(samples)
std_dev_val = np.std(samples)

print(f"Generated {len(samples)} samples.")
print(f"Sample Mean: {mean_val:.4f}")
print(f"Sample Std Dev: {std_dev_val:.4f}")

# Check how many values fall within 1 standard deviation of the mean
within_one_sigma = np.sum(np.abs(samples - mean_val) < std_dev_val)
percentage_within_one_sigma = (within_one_sigma / len(samples)) * 100

print(f"Percentage within 1 sigma: {percentage_within_one_sigma:.2f}%")

Explanation: We import numpy and call np.random.normal to create an array of 1,000 floats. We then compute the actual sample mean and standard deviation. Due to randomness, these will be close to 0 and 1 but not exact. Finally, we count how many points lie within one standard deviation of the mean. For a true normal distribution, approximately 68% of data falls within this range.

Common mistakes

• Confusing Variance and Standard Deviation: The scale parameter expects the standard deviation, not the variance. If you have a variance of 4, pass scale=2.
• Ignoring Sample Size: Small sample sizes (e.g., size=10) may produce means and standard deviations that differ significantly from the theoretical parameters due to high variance.
• Assuming All Data is Normal: Not all datasets are normally distributed. Always check skewness and kurtosis before applying statistical methods that assume normality.
• Forgetting Random Seed: Without setting np.random.seed(), results will vary every time you run the code, making debugging difficult.

When to use it

Compare the normal distribution with other common distributions to choose the right model.

Practice

Guided Exercise: Modify the example above to generate 5,000 samples with a mean of 50 and a standard deviation of 5. Print the percentage of samples that fall between 45 and 55.

Challenge: Generate two sets of 1,000 samples: one with scale=1 and one with scale=3. Plot histograms of both using matplotlib.pyplot.hist to visually compare their spreads.

Hint: For the guided exercise, use boolean indexing: np.sum((samples > 45) & (samples < 55)).

Quick check

Question: If you want to simulate test scores centered at 75 with a typical variation of 10 points, what arguments should you pass to np.random.normal?

Answer: You should pass loc=75 and scale=10.

Summary

The normal distribution is a fundamental concept in statistics characterized by its symmetric bell shape, defined by mean and standard deviation. Using np.random.normal allows you to simulate this behavior for data analysis, machine learning initialization, and statistical testing. Remember that while powerful, it is an assumption that must be validated against your specific dataset.