Normal Approximations

Hey students! 👋 Ready to dive into one of the most powerful tools in statistics? In this lesson, we'll explore how and when to use normal approximations for binomial and other distributions, including the crucial concept of continuity corrections. By the end of this lesson, you'll understand when it's appropriate to swap out complex binomial calculations for simpler normal distribution methods, and you'll master the art of applying continuity corrections to get accurate results. This skill will save you tons of time in exams and help you tackle real-world statistical problems with confidence! 🎯

Understanding When Normal Approximations Are Valid

The magic of normal approximations lies in knowing exactly when you can use them. For binomial distributions, there are specific conditions that must be met before you can safely make the switch to a normal approximation.

The golden rules for using normal approximations with binomial distributions are:

• np > 5 (where n is the number of trials and p is the probability of success)
• nq > 5 (where q = 1 - p, the probability of failure)
• The sample size n should be reasonably large (typically n ≥ 30 is preferred)

Let's see why these conditions matter! 📊 Imagine you're flipping a fair coin 100 times. Here, n = 100, p = 0.5, and q = 0.5. We get np = 100 × 0.5 = 50 and nq = 100 × 0.5 = 50. Both values are much greater than 5, so we're good to go!

But what if you're looking at a rare event? Say you're studying a manufacturing process where only 2% of items are defective, and you're checking 20 items. Here, n = 20, p = 0.02, and q = 0.98. We get np = 20 × 0.02 = 0.4 and nq = 20 × 0.98 = 19.6. Since np < 5, the normal approximation wouldn't be reliable here - the binomial distribution is too skewed to be approximated well by the symmetric normal curve.

When these conditions are met, the binomial distribution B(n, p) can be approximated by the normal distribution N(μ, σ²), where:

• μ = np (the mean)
• σ² = npq (the variance)
• σ = √(npq) (the standard deviation)

The Continuity Correction: Bridging Discrete and Continuous

Here's where things get really interesting! 🌉 The binomial distribution is discrete - it only takes whole number values (you can't have 2.5 successes in coin flips). But the normal distribution is continuous - it can take any real value. This creates a problem when we try to approximate one with the other.

Enter the continuity correction - a simple but crucial adjustment that accounts for this difference. Think of it as building a bridge between the discrete and continuous worlds.

Here's how it works: when you want to find P(X = k) for a discrete distribution, you approximate it using P(k - 0.5 < Y < k + 0.5) for the continuous normal distribution. The 0.5 adjustment is the continuity correction.

Let's break down the different scenarios:

For P(X = k): Use P(k - 0.5 < Y < k + 0.5)

For P(X ≤ k): Use P(Y < k + 0.5)

For P(X ≥ k): Use P(Y > k - 0.5)

For P(X < k): Use P(Y < k - 0.5)

For P(X > k): Use P(Y > k + 0.5)

Why does this work? Picture the discrete values as bars in a histogram. Each discrete value k represents a bar centered at k, extending from k - 0.5 to k + 0.5. The continuity correction ensures we're capturing the right area under the normal curve! 📈

Real-World Example: Quality Control in Manufacturing

Let's put this into practice with a real scenario! 🏭 Imagine you work for a smartphone manufacturer, and historical data shows that 8% of phones fail a particular quality test. Your supervisor asks you to find the probability that in a batch of 200 phones, exactly 15 will fail the test.

First, let's check if we can use the normal approximation:

• n = 200, p = 0.08, q = 0.92
• np = 200 × 0.08 = 16 > 5 ✓
• nq = 200 × 0.92 = 184 > 5 ✓

Great! We can use the normal approximation. The binomial distribution B(200, 0.08) can be approximated by N(16, 14.72), since:

$- μ = np = 16$

• σ² = npq = 200 × 0.08 × 0.92 = 14.72

$- σ = √14.72 ≈ 3.84$

Now, to find P(X = 15), we use the continuity correction:

P(X = 15) ≈ P(14.5 < Y < 15.5)

Converting to standard normal: Z = (Y - μ)/σ

P(14.5 < Y < 15.5) = P((14.5 - 16)/3.84 < Z < (15.5 - 16)/3.84)

$= P(-0.39 < Z < -0.13)$

Using standard normal tables, this gives us approximately 0.099 or about 9.9%. Without the continuity correction, our answer would be less accurate!

Extending to Other Distributions

Normal approximations aren't just for binomial distributions! 🚀 The Poisson distribution can also be approximated by a normal distribution when the parameter λ (lambda) is large, typically when λ > 5.

For a Poisson distribution with parameter λ, the normal approximation is N(λ, λ), since both the mean and variance of a Poisson distribution equal λ.

Consider this example: A busy coffee shop serves an average of 50 customers per hour, following a Poisson distribution. What's the probability they serve between 45 and 55 customers in a given hour?

Since λ = 50 > 5, we can use the normal approximation N(50, 50), so σ = √50 ≈ 7.07.

With continuity correction:

P(45 ≤ X ≤ 55) ≈ P(44.5 < Y < 55.5)

This approach makes complex Poisson calculations much more manageable! ☕

Common Pitfalls and How to Avoid Them

Watch out for these common mistakes! ⚠️

Mistake 1: Forgetting to check the conditions before applying normal approximations. Always verify np > 5 and nq > 5 for binomial distributions!

Mistake 2: Applying continuity corrections incorrectly. Remember: for "less than or equal to" (≤), add 0.5; for "greater than or equal to" (≥), subtract 0.5.

Mistake 3: Using normal approximations when the sample size is too small or when p is too close to 0 or 1. When p < 0.1 or p > 0.9, you need larger sample sizes for the approximation to work well.

Mistake 4: Forgetting that the normal approximation is just that - an approximation! It's most accurate near the center of the distribution and less accurate in the tails.

Conclusion

Normal approximations are incredibly powerful tools that transform complex discrete probability calculations into manageable continuous ones! Remember the key conditions: np > 5 and nq > 5 for binomial distributions, and λ > 5 for Poisson distributions. Always apply continuity corrections when moving from discrete to continuous, and double-check your work by ensuring your conditions are met. With practice, you'll find these approximations save you significant time while maintaining accuracy in your statistical analyses.

Study Notes

• Normal approximation conditions for binomial B(n,p):

• np > 5 and nq > 5 (where q = 1-p)
• Approximated by N(np, npq)

• Continuity correction rules:

• P(X = k) → P(k - 0.5 < Y < k + 0.5)
• P(X ≤ k) → P(Y < k + 0.5)
• P(X ≥ k) → P(Y > k - 0.5)
• P(X < k) → P(Y < k - 0.5)
• P(X > k) → P(Y > k + 0.5)

• Binomial to normal parameters:

$ - Mean: μ = np$

$ - Variance: σ² = npq$

• Standard deviation: σ = √(npq)

• Poisson approximation condition:

• λ > 5, approximated by N(λ, λ)

• Key formula for standardizing: Z = (Y - μ)/σ

• Always check conditions before applying approximations

• Continuity correction accounts for discrete → continuous conversion

• Most accurate near the center, less accurate in distribution tails