Scribd
Upload
81 views4 pages

Math 362, Problem Set 5

This document contains a math problem set that is due on March 16, 2010. It includes 6 problems related to statistics, random variables, and probability distributions. The problems cover topics like chi-square tests, generating random variables from probability distributions, bootstrap sampling, confidence intervals, and expectations and variances of random variables.

Uploaded by

toancao
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
81 views4 pages

Math 362, Problem Set 5

This document contains a math problem set that is due on March 16, 2010. It includes 6 problems related to statistics, random variables, and probability distributions. The problems cover topics like chi-square tests, generating random variables from probability distributions, bootstrap sampling, confidence intervals, and expectations and variances of random variables.

Uploaded by

toancao
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Math 362, Problem set 5

Due 3/16/10

1. (3.7.6) Another estimating chi-square: Let the result of a random exper-


iment be classifies as one of the mutually exclusive and exhaustive ways
A1 , A2 , A3 and also as one of the mutually exclusive and exhaustive ways
B1 , B2 , B3 , B4 . Two hundred independent trials of the experiment result
in the following data

B1 B2 B3 B4
A1 10 21 15 6
A2 11 27 21 13
A3 6 19 27 24

Test, at the 0.05 significance level the hypothesis of independence of the


A and B attribute, namely H0 : P(Ai ∩ Bi ) = P(Ai )P(Bi ), i = 1, 2, 3 and
j = 1, 2, 3, 4 against the alternative of dependence.
Answer:
10+21+15+6 13
We estimate P(A1 ) = 200 = 50 . Likewise we estimate:

18 19 27
P(A2 ) = P(A3 ) = P(B1 ) =
50 50 200
67 63 43
P(B2 ) = P(B3 ) = P(B4 ) =
200 200 200
Note that, really, we’ve only estimated 5 probabilities, because P(A3 ) is
determined once P(A2 ) and P(A1 ) is estimated, and likewise for P(B4 ).
Let nij denote the number of events in Ai ∩Bj (so n23 = 21). We compute

X (nij − 200P(Ai )P(Bj )2


≈ 12.941.
200P(Ai )P(Bj )
Since we estimated 5 parameters, we compare to a χ2 (12 − 1 − 5) = χ2 (6)
random variable. Since 12.941 > 12.592, we reject H0 .
2. (5.8.5) Determine a method to generate random observations for the fol-
lowing pdf: f (x) = 4x3 for 0 < x < 1, zero elsewhere.

1
3. (5.8.18) For α > 0 and β > 0, consider the following accept/reject algo-
rithm:
(1) Generate U1 and U2 iid uniform(0,1) random variables. Set V1 =
1/α 1/β
U1 and V2 = U2 .
(2) Set W = V1 + V2 . If W ≤ 1, set X = V1 /W , else goto Step(1).
(3) Deliver X.
Show that X has a beta distribution with parameters α and β.
Note/Hint: The analysis is quite similar to the analysis of the algorithm
we did in class. That is, look for the cdf P(X ≤ x) and note that it is
some conditional probability.
Answer: Our goal is to show that fX (x) = Cxα−1 (1 − x)β−1 for some
constant C. We note that

FX (x) = P(X ≤ x) = P(V1 /W ≤ x|W ≤ 1)


P(V1 ≤ xW, W ≤ 1)
=
P(W ≤ 1)
= cP(V1 ≤ xW, W ≤ 1),

for some constant C. We find the joint distribution of V1 and W as


follows. First we find that fV1 (v1 ) = αv1α−1 for 0 < v1 < 1, and likewise
fV2 (v2 ) = αv1α−1 . Noting that V2 = W − V1 , we have that the Jacobian
of the transformation V = V1 and W = V1 + V2 is 1. Thus the joint
distribution of fV1 ,W (v, w) = αβv α−1 (w − v)β−1 for 0 < v < 1, and
v < w < v + 1. We have that
Z 1 Z wx
FX (x) = c αβv α−1 (w − v)β−1 dvdw
w=0 v=0

Taking the derivative and applying the fundamental theorem of calculus,


we have
Z 1 Z 1
α−1 β−1 α−1 β−1
fX (x) = c αβ(wx) (w−wx) dw = x (1−x) c αβwα+β−2 dw
w=0 w=0

Since this is of the proper form, the normalizing constant must be correct
and this is a beta distribution as desired.
4. (5.9.1) Let x1 , . . . , xn be the values of a random sample. A bootstrap
0
sample x∗ = (x∗1 , . . . , x∗n ) is a random sample of x1 , . . . , xn drawn with
replacement.

(a) Show that E[x∗i ] = x̄


(b) If n is odd, show that median {x∗i } = x((n+1)/2)
Pn
(c) Show that Var(x∗i ) = n1 i=1 (xi − x̄)2

2
Notes: There was a part (a) on the book problem, which I cut as it’s
obvious but hard to write up nicely. It asks you to show that the x∗i are
independent (which they are because I chose with replacement) and have
the CDF F̂n which is F̂n (x) = n1 × (# of xi < x), which is obvious because
I select the x∗i uniformly from x1 , . . . , xn . Note that in (c), you have n1
and not n−11
because x̄ is the real mean of the x∗i as opposed to the sample
mean of the x∗i .
Answer:

X 1X
E[x∗i ] = xi P(x∗i = xi ) = xi = x̄.
n
For (b), since half of the values are above, and half are below x(n+1)/2 ,
this is the median as desired.
For (c), we have
1X ∗
Var(x∗i ) = E[(x∗i − E[x∗i ])2 ] = (xi − x̄)2 .
n

5. (5.9.2) Let X1 , . . . , Xn be a random sample from a Γ(1, β) distribution.

(a.) Show that the confidence interval (2nX̄/(χ22n )(1−(α/2)) , 2nX̄/(χ22n )α/2 )
is an exact (1 − α)100% confidence interval for β.
Notation note: (χ22n )(1−(α/2)) is the books (rather confusing) way of
saying ’the number such that P(χ2 (2n) < (χ22n )(1−(α/2)) ) = 1 − α/2;
for instance for α = .5 and n = 4, so that 2n = 8, we have (by our
chart) (χ28 )(1−(α/2)) = 17.535. Unfortunately the notation makes the
problem look harder than it is.
(b.) Using part (a), show that the 90% confidence interval discussed in
Example 5.9.1 is (64.99, 136.69).
2nX̄
Answer: For (a) note that 2nX̄ has a Γ(n, 2β) distribution and β has
a Γ(n, 2) = χ2 (2n) distribution. Thus:

2nX̄
P((χ22n )α/2 ≤ ≤ (χ22 n)1−α/2 ) = 1 − α.
β
Solving the interval for β, we get the desired answer.
For (b), we get that (χ240 ).05 = 26.5 and (χ240 ).95 = 55.8 (sadly from the
internet). Then we compute to see the hypothesized interval.
6. (5.9.10) Let z ∗ be drawn at random from the discrete distribution which
has mass n−1 at each point zi = xi − x̄ + µ0 , where (x1 , . . . , xn ) is the
realization of a random sample. Determine E[z ∗ ] and Var(z ∗ ).
Answer:

3
1X
E[z ∗ ] = (xi − x̄ + µ0 ) = µ0 .
n
Likewise
1X 1X
Var(z ∗ ) = (xi − x̄ + µ0 − µ0 )2 = (xi − x̄)∗ = Var(x∗ ).
n n
where x∗ is as in problem 3.

You might also like