120k views
5 votes
2. The number of defects in a 400-metre roll of magnetic recording tape has a Poisson distribution with unknown parameter μ, which has a prior Gama distribution of the form μ-Ga(3,1). When five rolls of this tape are selected at random and inspected, the numbers of defects found on the rolls are 2, 2, 6, 0 and 3. x9-1 [probability density function of gamma is Ga(x, a, ß) = 0,ß > 0] Γ(α) -Ba e-Bx, x>0₁α > a) Determine expressions for the likelihood function and posterior probability density function of μ. (17 marks) b) Show that the posterior probability mass function of X given the data above is 616r(x + 16) P(μ\X) = 15! x! 7x+16 [Hints: P(u\X) = f(x,μ)ƒ (μ|X)dµ‚μ> 0 and [(x) = f tx-¹e-t dt] c) Given that the median of Beta distribution is m(a, ß) = a+ß- Find the Bayesian estimate of μ under the absolute error loss function.

User MikMik
by
8.2k points

1 Answer

3 votes

Answer:

(A)

f(μ|x) = (μ^(x1+x2+x3+x4+x5+2) * e^(-26μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))) / ∫_0^∞ μ^(x1+x2+x3+x4+x5+2) * e^(-26μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3)) dμ

(B)

3/1 = 3

Explanation:

a) The likelihood function of μ is the probability of observing the given data, given a particular value of μ. Since the number of defects in a 400-meter roll of magnetic recording tape has a Poisson distribution with parameter μ, the likelihood function can be expressed as follows:

L(μ|x) = P(X1 = x1, X2 = x2, X3 = x3, X4 = x4, X5 = x5 | μ)

= P(X1 = x1 | μ) * P(X2 = x2 | μ) * P(X3 = x3 | μ) * P(X4 = x4 | μ) * P(X5 = x5 | μ)

= e^(-5μ) * (μ^x1 / x1!) * e^(-5μ) * (μ^x2 / x2!) * e^(-5μ) * (μ^x3 / x3!) * e^(-5μ) * (μ^x4 / x4!) * e^(-5μ) * (μ^x5 / x5!)

= e^(-25μ) * (μ^(x1+x2+x3+x4+x5) / (x1! * x2! * x3! * x4! * x5!))

where x1 = 2, x2 = 2, x3 = 6, x4 = 0, and x5 = 3.

The posterior probability density function of μ can be obtained using Bayes' theorem. According to Bayes' theorem, the posterior probability density function of μ given the observed data x is proportional to the product of the likelihood function and the prior probability density function of μ:

f(μ|x) ∝ L(μ|x) * f(μ)

where f(μ) is the prior probability density function of μ, which is given as μ ~ Ga(3,1). Therefore,

f(μ) = μ^(3-1) * e^(-μ/1) / Γ(3) = μ^2 * e^(-μ)

Substituting the values of L(μ|x) and f(μ), we get

f(μ|x) ∝ e^(-25μ) * (μ^(x1+x2+x3+x4+x5) / (x1! * x2! * x3! * x4! * x5!)) * μ^2 * e^(-μ)

= μ^(x1+x2+x3+x4+x5+2) * e^(-26μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))

Thus, the posterior probability density function of μ given the observed data x is:

f(μ|x) = (μ^(x1+x2+x3+x4+x5+2) * e^(-26μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))) / ∫_0^∞ μ^(x1+x2+x3+x4+x5+2) * e^(-26μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3)) dμ

b) To find the posterior probability mass function of X given the data above, we can use the formula:

P(μ|X) = f(X|μ) * f(μ) / f(X)

where f(X|μ) is the Poisson probability mass function with parameter μ, f(μ) is the Gamma probability density function with parameters α = 3 and β = 1, and f(X) is the marginal probability mass function of X, which can be obtained by integrating the joint density function of X and μ over μ:

f(X) = ∫_0^∞ f(X|μ) * f(μ) dμ = ∫_0^∞ e^(-μ) * μ^(X+2) / (X! * Γ(3)) * μ^2 * e^(-μ) dμ

= Γ(X+3) / (X! * Γ(3))

where X = x1 + x2 + x3 + x4 + x5.

Therefore, we have:

P(μ|X) = f(X|μ) * f(μ) / f(X)

= e^(-5μ) * μ^x1 / x1! * e^(-5μ) * μ^x2 / x2! * e^(-5μ) * μ^x3 / x3! * e^(-5μ) * μ^x4 / x4! * e^(-5μ) * μ^x5 / x5! * μ^2 * e^(-μ) / ∫_0^∞ e^(-5μ) * μ^x1 / x1! * e^(-5μ) * μ^x2 / x2! * e^(-5μ) * μ^x3 / x3! * e^(-5μ) * μ^x4 / x4! * e^(-5μ) * μ^x5 / x5! * μ^2 * e^(-μ) dμ

= (μ^(x1+x2+x3+x4+x5+2) * e^(-55 - μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))) / ∫_0^∞ μ^(x1+x2+x3+x4+x5+2) * e^(-55 - μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3)) dμ

Simplifying the expression, we get:

P(μ|X) = (μ^(x1+x2+x3+x4+x5+2) * e^(-30 - μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))) / 616 * (μ^(x1+x2+x3+x4+x5+2) * e^(-30 - μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3))) dx

Therefore, the posterior probability mass function of X given the observed data is:

P(μ\X) = 616r(x + 16) * (μ^(x1+x2+x3+x4+x5+2) * e^(-30 - μ) / (x1! * x2! * x3! * x4! * x5! * Γ(3)))

c) The Bayesian estimate of μ under the absolute error loss function is given by:

μ_B = E[μ|X] = ∫_0^∞ μ * f(μ|X) dμ

To find the value of μ_B, we can use the fact that the Gamma distribution with parameters α and β has a median of m(α, β) = α/β. Therefore, we can choose the value of μ_B that minimizes the absolute difference between the median of the posterior distribution and the observed data:

|α/β - (x1+x2+x3+x4+x5+3)/31| = |3/1 - (2+2+6+0+3+3)/31| = 0.0645

Hence, the Bayesian estimate of μ under the absolute error loss function is 3/1 = 3.

User Kahn Kah
by
8.5k points