Normal approximation for binomial probability with n=14, p=0.5

2 replies · 3K views · last reply
g.lemaitre
Messages
267
Reaction score
2

Homework Statement



find the normal approximation for the binomial probability P(x = 4,5) where n=14 and p = .5.

Homework Equations



μ = np
σ = sqrt(npq)

z = (x - μ)/σ

The Attempt at a Solution



p = .5 q = .5

μ = 14*.5 = 7

σ = sqrt(14 * .5 * .5) = 1.87

z = (4 - 7)/1.87 = -1.61

(my book uses tables to convert the z score into the probability of getting x < 4)

z = -1.61 = .5 - .4463 = .0537

The book says the answer is .1812 which is not what I'm getting.


z =
 
it's rare that such a simple problem takes this long. let me provide the example from the book

Textbook page titled Solving a Binomial Probability Problem with the Normal Distribution, showing the solution to a firing pin inspection problem with n = 25 and p = 1/3, including the binomial formula and mean and standard deviation calculations.


Textbook page explaining normal approximation to a binomial probability: a bell curve with mean 8.333 and standard deviation 2.357 shaded left of 3.5, with z = -2.05 and P(x less than 3.5) = 0.0202.
 
g.lemaitre said:
it's rare that such a simple problem takes this long. let me provide the example from the book

Textbook page titled Solving a Binomial Probability Problem with the Normal Distribution, showing the solution with formula P(x) for x = 0 to 25, and mean and standard deviation calculations: μ = np = 8.333 and σ = 2.357.


Textbook page excerpt: a bell-shaped normal curve with mean 8.333 and standard deviation 2.357 marked, with the shaded left tail up to x = 3.5 on the axis labelled 0 1 2 3, then P(x is no more than 3) = P(x ≤ 3) = P(x < 3.5), z = (3.5 - 8.333)/2.357 = -2.05, and P(z < -2.05) = 0.5000 - 0.4798 = 0.0202.

The normal approximation is not very good in this example (because N = 25 is not very large and z is more than 2 standard deviations below the mean). P_exact = 0.0148904, while the continuity-corrected normal approximation is about 0.020152 (so using the normal gives about a 35% error). The normal approximation would be better if N were larger or z were closer to the mean.

RGV