STAT0005: Probability and Inference Level 5, 2018/19
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
STAT0005: Probability and Inference Level 5, 2018/19
Answer ALL questions. Section A carries 40% of the total marks and Section B carries 60%. The relative weights attached to each question are as follows: A1(10), A2(11), A3(9), A4(10), B1(31), B2(29). The numbers in square brackets indicate the relative weight attached to each part question.
You may use without proof the following results:
❼ A random variable X follows the Poisson distribution with parameter
µ > 0, denoted X ~ Poi(µ), if and only if it has pmf
P (X = k) = eo. , k e {0, 1, 2, 3, . . .}.
In this case E[X] = µ, Var(X) = µ and Gx (z) = exp (µ(z _ 1)).
❼ A random variable X follows the Geometric distribution with parame-
ter p e (0, 1), denoted X ~ Geo(p), if and only if it has pmf
P (X = k) = (1 _ p)k o1p, k e N.
1 1 _ p
p p2 .
❼ A random variable X follows the Binomial distribution with parameters
n e N and p e (0, 1), denoted X ~ Bin(n, p), if and only if it has pmf
P (X = k) = ╱ 、k(n) pk (1 _ p)nok , k e {0, 1, 2, . . . , n}.
In this case E[X] = np and Var(X) = np(1 _ p).
❼ A random variable X follows the Gamma distribution with parameters
α > 0 and λ > 0 if it has pdf
,
fx (x) = .
λ9 x9o1 eoAz |
Γ(α) 0 |
if x > 0
otherwise .
In this case E[X] = , Var(X) = and Mx (s) = ╱ 1 _ 、o9 for
s < λ. The Gamma function is given by Γ(s) = 0l xso1 eoz dx; if s e N then we have Γ(s) = (s _ 1)!. Finally, if X ~ Gam(α, λ) and Y ~ Gam(β, λ) are two independent random variables (with α, β, λ > 0) then X + Y ~ Gam(α + β, λ) follows.
Section A
A1 Let X follow the distribution given by the following pmf:
k 0 1 2 3
P (X = k) 1/4 1/4 1/4 1/4
(a) Compute the cdf and sketch it bearing in mind that it is defined on the whole real axis. [6]
(b) Consider the transformed random variable Y = |X _ 2|. Compute its pmf, name the distribution and provide any parameter value(s) required as well as the expected value and variance. [4]
A2 Let X ~ Geo(p) with p e (0, 1). Conditionally on X , Y takes the value
X with probability 1/2 and the value 3X with probability 1/2.
(a) Compute EY ax [Y |X].
(b) Compute E[Y].
(c) Compute VarY ax(Y |X).
(d) Show that Var(Y) = 6po2 _ 5po1 .
[2]
[3]
[2]
[4]
µk
k!(e. _ 1)
e.z _ 1
(a) Show that Gx (z) =
(b) Hence or otherwise compute E[X].
[5]
[4]
A4 Let X1 and X2 denote two tosses of a potentially unfair coin, such that
P (X1 = 1) = P (X2 = 1) = p and P (X1 = 0) = P (X2 = 0) = 1 _ p for some p e (0, 1). In this question, all coin tosses are independent.
(a) Let W = (X1 _ X2 )2 . Compute the pmf of W .
(b) Compute P (X1 = 1|W = 1) and P (X1 = 0|W = 1).
[3]
[4]
(c) Let Y denote the number of replications of the above two-toss experiment until W = 1 happens. Note that Y takes values in N. What is its distribution including any parameter value(s) needed?
[3]
Section B
B1 For n e N, let X1 , X2 , . . . , Xn be i.i.d. random variables each with pdf
f(x) = eoz如 /θ for x e (0, o),
where θ e (0, o) is an unknown parameter.
(a) Show that = Xi(2) is the MLE of θ. [7] (b) Show that the sampling distribution of is Gam(n, nθ o1 ). 血〕]n) Compute the distribution of X1(2) first. You may also use the ad- ditive property of Gamma-distributed random variables given in
the preamble.
(c) Hence or otherwise compute the bias and variance of
[6]
[2]
(d) Compute the CRLB for θ. Does achieve the same mse as an unbiased estimator attaining the CRLB? Justify. [5]
(e) Obtain an approximate 95% confidence interval for θ based on
asymptotic normality of [3]
(f) Remembering that an interval [a(, n), b(, n)] is a 95% confidence interval if and only if P ([a(, n), b(, n)] 3 θ) > 0.95, decide whether the interval computed in (e) actually is a 95% confidence interval in the case n = 2. Discuss your findings commenting on the statistical implications of using the approximate confidence in- terval in (e) assessing their severity. 血〕]n) Using the sampling distribution of established in (b), you will need to calculate the
probability P ╱ < 、. You may require integration
by parts and a pocket calculator. [8]
B2 (a) State the definition of the mgf Mx of a random variable X. [3]
(b) State how the pgf Gx and the mgf Mx of a discrete random variable X taking values in N are related. [3]
(c) What value does Mx (s) take at s = 0? Does this depend on the distribution of X? Justify your answer. [4]
(d) Explain carefully how mgfs are used to establish independence and how they are used to decide whether two random variables follow the same distribution. Your explanation should comment on why it is important to include the range of s for which Mx (s) exists when reporting the mgf Mx . [7]
(e) Fix k e N and λ > 0. For this question you may use the following fact without proof:
For i e N, let X N(-λ , 1). Then, Y = i(k)=1 Xi(2) has mgf
MY (s) = (1 _ 2s)ok/2 exp ╱ 、 , s e (_o, 1/2).
(i) For i e N, let Z χ1(2). For j e N, compute the mgf of U = Zi . Name the distribution of U and provide any parameter value(s) needed. [6]
(ii) Now let J ~ Poi(kλ/2) be independent of the Zi . Compute the mgf of W = Zi , decide whether W and Y follow the same distribution and justify your decision. 血〕]n) Con- ditioning may help you re-use previous results. [6]
2022-04-19