This is a beta course, so its structure, chapters, and examples may continue to change.
2016 National Graduate Entrance Examination Mathematics I Past Paper
Part I: Multiple Choice
Part I: Multiple Choice: Questions 1-8, 4 points each, 32 points in total.
1. If the improper integral ∫0+∞xa(1+x)b1dx converges, then ( )
(A) a<1 and b>1
(B) a>1 and b>1
(C) a<1 and a+b>1
(D) a>1 and a+b>1
Improper Integral
Approach:
This question tests the determination of the convergence or divergence of an improper integral with parameters. Both ends of the interval of integration are improper points, so the order of the integrand relative to the corresponding power function has to be examined separately at x→0+ and at x→+∞.
Steps:
At x→0+, (1+x)b→1, so the integrand is of the same order as x−a, and the convergence of ∫01x−adx requires a<1.
At x→+∞, xa(1+x)b∼xa+b, so the integrand is of the same order as x−(a+b), and the convergence of ∫1+∞x−(a+b)dx requires a+b>1.
The improper integral converges only if it converges at both ends, so the necessary and sufficient condition is a<1 and a+b>1, and option (C) is correct.
Answer:
C.
2. Given the function f(x)={2(x−1),lnx,x<1x⩾1, one antiderivative of f(x) is ( )
This question tests how to find an antiderivative of a piecewise function. A candidate must be differentiable everywhere: the derivative must match f on each piece, and at the junction point x=1 the function must be continuous with matching left and right derivatives.
Steps:
For x<1 we have f(x)=2(x−1), so an antiderivative on that piece is ∫2(x−1)dx=(x−1)2+C1; for x⩾1 we have f(x)=lnx, so it is ∫lnxdx=x(lnx−1)+C2. All four options agree on the piece x<1, so only the piece x⩾1 needs checking.
An antiderivative must be continuous at x=1. For (A) the right-hand piece has value 1⋅(ln1−1)+0=−1 at x=1, while the left-hand limit is (1−1)2+0=0; the two differ, so (A) is not continuous.
For (B) the right-hand piece has value 1⋅(ln1+1)−1=0 at x=1, matching the left-hand limit 0, so (B) is continuous; for (C) the right-hand value is 1⋅(ln1+1)+1=2=0, so (C) is not continuous; for (D) the right-hand value is 1⋅(ln1−1)+1=0, matching 0, so (D) is continuous.
Now check differentiability. For (B): on x>1 we have F′(x)=lnx+1−1=lnx=f(x), and on x<1 we have F′(x)=2(x−1)=f(x); at x=1,
limx→1−x−1F(x)−F(1)=limx→1−(x−1)=0limx→1+x−1F(x)−F(1)=limx→1+x−1x(lnx+1)−1=L’Hopitallimx→1+1lnx+1=1
The one-sided derivatives differ, so (B) is not differentiable and is not an antiderivative.
For (D): on x>1 we have F′(x)=lnx−1+1=lnx=f(x), and on x<1 we have F′(x)=2(x−1)=f(x); at x=1,
limx→1−x−1F(x)−F(1)=limx→1−(x−1)=0,limx→1+x−1F(x)−F(1)=limx→1+x−1x(lnx−1)+1=L’Hopitallimx→1+lnx=0
Both one-sided derivatives exist and equal 0=f(1), so (D) is an antiderivative of f.
Answer:
D.
3. If y=(1+x2)2−1+x2 and y=(1+x2)2+1+x2 are two solutions of the equation y′+p(x)y=q(x), then q(x)= ( )
(A) 3x(1+x2)
(B) −3x(1+x2)
(C) 1+x2x
(D) −1+x2x
First-Order Linear ODE
Approach:
This question tests the structural properties of the solutions of a first-order linear differential equation. Two solutions are given, and their difference must be a solution of the corresponding homogeneous equation, so p(x) can first be determined; then one particular solution is substituted back into the equation to determine q(x).
Steps:
The difference of the two solutions 21+x2 is a nonzero solution of the homogeneous equation y′+p(x)y=0; from 1+x2=e−∫p(x)dx we get ∫p(x)dx=−21ln(1+x2), that is p(x)=−1+x2x.
Substitute y=(1+x2)2+1+x2 and its derivative y′=4x(1+x2)+1+x2x into the equation y′+p(x)y=q(x).
Combining term by term gives q(x)=4x(1+x2)+1+x2x−x(1+x2)−1+x2x=3x(1+x2), so option (A) is correct.
Answer:
A.
4. Given the function f(x)=⎩⎨⎧x,n1,x⩽0n+11<x⩽n1,n=1,2,⋯, then ( )
(A) x=0 is a first-kind discontinuity of f(x)
(B) x=0 is a second-kind discontinuity of f(x)
(C) f(x) is continuous but not differentiable at x=0
(D) f(x) is differentiable at x=0
discontinuitycontinuitydifferentiability
Approach:
This question tests continuity and differentiability of a piecewise function at its junction point. Compute the one-sided limits and the value of the function to decide continuity, then use the difference quotient to decide differentiability.
Steps:
For x⩽0 we have f(x)=x, so x→0−limf(x)=0, and f(0)=0.
For x>0 we have f(x)=n1 with n+11<x⩽n1. As x→0+ we have n→∞, hence x→0+limf(x)=n→∞limn1=0.
Both one-sided limits exist and equal f(0)=0, so f(x) is continuous at x=0; neither (A) nor (B) holds.
Now check the derivative. The left derivative is
f−′(0)=limx→0−xf(x)−f(0)=limx→0−xx=1
The right derivative does not exist. For any n, when x∈(n+11,n1] we have xf(x)−f(0)=nx1, whose values grow with n (taking x=n1 gives the difference quotient 1, while taking x=2n+21 gives approximately 2), so the right-hand limit fails to exist.
Therefore f(x) is continuous but not differentiable at x=0, so (C) is the correct choice.
Answer:
C.
5. Suppose A,B are invertible and similar; which of the following is wrong ( )
(A) AT and BT are similar
(B) A−1 and B−1 are similar
(C) A+AT and B+BT are similar
(D) A+A−1 and B+B−1 are similar
Matrix Similarity
Approach:
This question tests the properties of similar matrices. A similarity transformation preserves the transpose, the inverse, and the polynomial combinations built out of them; examine each of the four options in turn to see which conclusion is preserved and which is not, and pick out the wrong one.
Steps:
Write B=P−1AP (P invertible); then BT=PTAP−T, that is AT and BT are similar, so option (A) is correct.
Similarly B−1=P−1A−1P, so A−1 and B−1 are similar, and option (B) is correct.
For option (D): B+B−1=P−1AP+P−1A−1P=P−1(A+A−1)P, so A+A−1 and B+B−1 are similar, and option (D) is correct.
For option (C): B+BT=P−1AP+PTAP−T, which in general cannot be written as P−1(A+AT)P, so A+AT and B+BT need not be similar, and option (C) is wrong.
Answer:
C.
6. For the quadratic form f(x1,x2,x3)=x12+x22+x32+4x1x2+4x1x3+4x2x3, the surface represented by f=2 is ( )
(A) an ellipsoid (B) a hyperboloid (C) a paraboloid (D) a cylinder
Quadric Surface
Approach:
This question tests the use of the standard form of a quadratic form to identify the type of a quadric surface. The key is to use the eigenvalues of the matrix of the quadratic form to determine the number of positive and negative signs in the standard form, and hence to decide whether the surface is an ellipsoid, a hyperboloid, a paraboloid or a cylinder.
Steps:
The matrix of the quadratic form f=x12+x22+x32+4x1x2+4x1x3+4x2x3 is A=122212221.
The eigenvalues of A are found to be 5,−1,−1, and after an orthogonal transformation the quadratic form becomes the standard form 5y12−y22−y32.
Setting f=2 gives 5y12−y22−y32=2; the standard form has both positive and negative signs and no zero coefficient, so it represents a hyperboloid, and option (B) is correct.
Answer:
B.
7. Let the random variable X∼N(μ,σ2)(σ>0), and write p=P{X⩽μ+σ2}. Then ( )
(A) p increases as μ increases
(B) p increases as σ increases
(C) p decreases as μ increases
(D) p decreases as σ increases
normal distributionstandard normal distribution function
Approach:
This question tests the standardisation of a normal random variable. Once the threshold is standardised, p turns out to depend only on σ and not on μ.
Steps:
Since X∼N(μ,σ2), we have σX−μ∼N(0,1); denote the standard normal distribution function by Φ.
Then
p=P{X⩽μ+σ2}=P{σX−μ⩽σσ2}=P{σX−μ⩽σ}=Φ(σ)
The expression for p involves σ only and is independent of μ, so neither (A) nor (C) holds.
Since Φ is strictly increasing on R, the larger σ is, the larger Φ(σ) is; that is, p increases as σ increases. Hence (B) is correct and (D) is not.
Answer:
B.
8. A random experiment E has three mutually exclusive outcomes A1,A2,A3, each occurring with probability 31. The experiment E is repeated independently twice; X is the number of times outcome A1 occurs and Y is the number of times outcome A2 occurs. Then the correlation coefficient of X and Y is ( )
(A) −21
(B) −31
(C) 31
(D) 21
correlation coefficientrepeated independent trialsbinomial distribution
Approach:
This question tests the correlation coefficient of two counts arising from repeated independent trials. Both X and Y follow the binomial distribution B(2,31), so by symmetry they have the same expectation and variance; the covariance follows from E(XY)−E(X)E(Y), where the key is to count the arrangements producing XY=1.
Steps:
Both X and Y follow the binomial distribution B(2,31), so
E(X)=E(Y)=2⋅31=32,D(X)=D(Y)=2⋅31⋅32=94
Compute E(XY). The product XY counts the times outcome A1 and outcome A2 occur across the two trials. Only a trial pair consisting of one A1 and one A2 gives XY=1; there are 2 such orderings, each with probability 31⋅31=91, hence
E(XY)=2⋅91=92
The correlation coefficient is
ρXY=D(X)D(Y)Cov(X,Y)=32⋅32−92=−21
So X and Y are negatively correlated (the two counts exclude each other within one trial), and (A) is the correct choice.
Answer:
A.
Part II: Fill in the Blank
Part II: Fill in the Blank: Questions 9-14, 4 points each, 24 points in total.
9. x→0lim1−cosx2∫0xtln(1+tsint)dt=______.
Limit
Approach:
This question tests the replacement by equivalent infinitesimals and the asymptotic estimate of an integral. The numerator is an integral with a variable upper limit, so one first has to find the leading term of the integral and then take the quotient with the equivalent expression for the denominator.
Steps:
As x→0 we have tsint∼t2, hence ln(1+tsint)∼tsint∼t2, and the integrand satisfies tln(1+tsint)∼t3.
Hence ∫0xtln(1+tsint)dt∼∫0xt3dt=4x4.
Also 1−cosx2∼2x4, so the original limit is x4/2x4/4=21.
Answer:
21.
10. The divergence of the vector field A(x,y,z)=(x+y+z)i+xyj+zk is divA=______.
Divergence
Approach:
This question tests the computation of the divergence of a vector field. The divergence is the sum of the partial derivatives of the components with respect to their own variables, so differentiate term by term according to the definition and then add.
Steps:
Write the vector field as A=(x+y+z,xy,z); the three components are P=x+y+z, Q=xy and R=z.
Differentiate term by term: ∂x∂P=1, ∂y∂Q=x, ∂z∂R=1.
Adding them gives divA=1+x+1=2+x.
Answer:
2+x.
11. Let f(u,v) be differentiable and let z=z(x,y) be determined by the equation (x+1)z−y2=x2f(x−z,y). Then dz∣(0,1)=______.
total differentialimplicit differentiationpartial derivative
Approach:
This question tests the computation of a total differential for an implicitly defined function. Write the equation as F(x,y,z)=0, differentiate with respect to x and y separately to obtain zx,zy, and then write the total differential.
Steps:
Put F(x,y,z)=(x+1)z−y2−x2f(x−z,y)=0; then F(x,y,z)=0 determines an implicit function z=z(x,y).
First find the value of z: substituting x=0,y=1 into the equation gives z−1=0, so at the point (0,1) we have z=1.
Differentiate with respect to x (noting that ∂x∂(x−z)=1−zx):
(x+1)zx+z−2xf(x−z,y)−x2fx(x−z,y)(1−zx)=0
At (0,1,z=1) the terms involving f vanish because x=0, and we obtain zx=0.
Differentiate with respect to y:
(x+1)zy−2y−x2fy(x−z,y)(1−zy)=0
At (0,1,z=1) this gives zy−2=0, i.e. zy=2.
Hence at the point (0,1)dz=zxdx+zydy=−dx+2dy
Answer:
−dx+2dy.
12. Let f(x)=arctanx−1+ax2x, and let f′′′(0)=1; then a=______.
Taylor Expansion
Approach:
This question tests the use of a Taylor expansion to compute a higher-order derivative. f′′′(0) is determined only by the coefficient of the cubic term of f(x) at 0, so it is enough to expand both parts up to the x3 term.
Steps:
Expand arctanx=x−3x3+o(x3).
Expand 1+ax2x=x(1−ax2+o(x2))=x−ax3+o(x3).
Subtracting gives f(x)=(a−31)x3+o(x3)=32a−1x3+o(x3), so f′′′(0)=3!⋅32a−1=2(2a−1)=1, and solving gives a=21.
Answer:
21.
13. The determinant λ004−1λ030−1λ200−1λ+1=______.
determinantcofactor expansion
Approach:
This question tests a structured computation of a high-order determinant. The last three rows of the determinant form an upper-shift structure, so expanding along the first column and then reducing the order is the most direct route.
Steps:
Denote the 4th-order determinant by D(λ). Expanding along the first column,
D(λ)=λ⋅M11−4⋅M41
The first minor is
M11=−1λ00−1λ00−1=(−1)3=−1
The second minor is
M41=−1λ0−1−1λ00−1=λ[λ(λ+1)+1]
Hence
D(λ)=−λ[λ2+λ+1]+4[λ2+λ+1]=λ4+λ3+2λ2+3λ+4
Numerical check: D(0)=4, D(1)=11 and D(2)=42, all agreeing with the formula above.
Answer:
λ4+λ3+2λ2+3λ+4.
14. Let X1,X2,⋯,Xn be a simple random sample from the population N(μ,σ2), with sample mean Xˉ=9.5. The upper confidence limit of the two-sided confidence interval for μ at confidence level 0.95 is 10.8. Then the two-sided confidence interval for μ at confidence level 0.95 is ______.
normal populationconfidence intervalsample mean
Approach:
This question tests the symmetry of the confidence interval for the mean of a normal population. Since the interval for μ is symmetric about the sample mean Xˉ, knowing the upper limit determines the lower limit.
Steps:
For a simple random sample from N(μ,σ2), the two-sided confidence interval for μ at level 1−α is
(Xˉ−uα/2nσ,Xˉ+uα/2nσ)
where uα/2 is the upper α/2 quantile of the standard normal distribution. This interval is symmetric about Xˉ.
The upper confidence limit is 10.8, so
Xˉ+u0.025nσ=10.8
With Xˉ=9.5 the lower confidence limit is
Xˉ−u0.025nσ=2Xˉ−10.8=19−10.8=8.2
Hence the required confidence interval is (8.2,10.8).
Answer:
(8.2,10.8).
Part III: Free Response
Part III: Free Response: Questions 15-23, 94 points in total.
15. (10 points) Given the plane region D={(r,θ)2⩽r⩽2(1+cosθ),−2π⩽θ⩽2π}, compute the double integral ∬Dxdxdy.
double integralpolar coordinates
Approach:
This question tests a double integral in polar coordinates. The inner and outer boundaries of D are the circle r=2 and the cardioid r=2(1+cosθ), the integrand is x=rcosθ, and the area element is dxdy=rdrdθ.
Steps:
Make the polar substitution x=rcosθ, y=rsinθ; then dxdy=rdrdθ and the integrand x becomes rcosθ.
Thus
∬Dxdxdy=∫−2π2πdθ∫22(1+cosθ)r2cosθdr
Integrate with respect to r:
∫22(1+cosθ)r2cosθdr=3cosθ[23(1+cosθ)3−23]=38cosθ[(1+cosθ)3−1]
Hence
∬Dxdxdy=38∫−2π2πcosθ[(1+cosθ)3−1]dθ
Expand (1+cosθ)3−1=3cosθ+3cos2θ+cos3θ. The terms cosθcos3θ=cos3θcosθ and cosθcos2θ=cos3θ are odd functions of θ, so they integrate to zero on the symmetric interval. What remains is
3∫−2π2πcos2θdθ=3⋅2π=23π
Therefore
∬Dxdxdy=38⋅23π=4π
Answer:
4π.
16. (10 points) Suppose the function y(x) satisfies y′′+2y′+ky=0, where 0<k<1.
(1) Prove that the improper integral ∫0+∞y(x)dx converges;
(2) If y(0)=1 and y′(0)=1, find ∫0+∞y(x)dx.
linear differential equation with constant coefficientsimproper integralcharacteristic equation
Approach:
This question tests a second-order linear homogeneous differential equation with constant coefficients together with an improper integral. The signs of the characteristic roots determine the exponential decay of y(x) and hence the convergence of the improper integral; for the second part, integrating the differential equation over [0,+∞) turns the required integral into endpoint values.
Steps:
(1) The characteristic equation is λ2+2λ+k=0, whose roots are
λ1=−1+1−k,λ2=−1−1−k
Since 0<k<1 we have 0<1−k<1, so λ1<0 and λ2<0.
The general solution is y(x)=C1eλ1x+C2eλ2x, and the improper integrals of both exponentials on [0,+∞) converge:
∫0+∞eλixdx=−λi1(i=1,2)
By linearity,
∫0+∞y(x)dx=C1∫0+∞eλ1xdx+C2∫0+∞eλ2xdx
converges, which proves the claim.
(2) Method one: since λ1,λ2<0,
limx→+∞y(x)=0,limx→+∞y′(x)=0
Rewrite the differential equation as y′′+2y′+ky=0 and integrate it over [0,+∞) (by integration by parts):
∫0+∞y′′(x)dx+2∫0+∞y′(x)dx+k∫0+∞y(x)dx=0
Here
∫0+∞y′′(x)dx=y′(x)∣0+∞=−y′(0)=−1,∫0+∞y′(x)dx=y(x)∣0+∞=−y(0)=−1
Method two: substituting y(0)=1,y′(0)=1 into the general solution gives
{C1+C2=1,λ1C1+λ2C2=1
Solving yields C1=21−k1(1−k+1) and C2=21−k1(1−k−1), and substituting these into
∫0+∞y(x)dx=−λ1C1−λ2C2=k3
agrees with the result above.
Answer:
(1) ∫0+∞y(x)dx converges;
(2) ∫0+∞y(x)dx=k3.
17. (10 points) Suppose the function f(x,y) satisfies ∂x∂f=(2x+1)e2xy and f(0,y)=y+1, and let L be a smooth curve from (0,0) to (1,t). Compute I(t)=∫L∂x∂fdx+∂y∂fdy, and find the minimum value of I(t).
Path-Independent Line Integral
Approach:
This question tests the total differential and path-independent line integrals. First integrate ∂f/∂x to obtain f and use the initial condition to fix the function of one unknown, then verify that the field is a gradient field so that the line integral is path independent, and finally evaluate it along a broken-line path and find its extreme value.
Steps:
With respect to x, integrating ∂x∂f gives f(x,y)=xe2xy+C(y); substituting the initial condition f(0,y)=y+1 gives C(y)=y+1, so f(x,y)=xe2xy+y+1.
Since ∂f/∂xdy=∂f/∂ydx, the vector field is a gradient field and the line integral is path independent, so evaluate it along the broken line from (0,0) to (1,0) and then to (1,t).
Adding the contributions along the two segments gives I(t)=21e2t+t; then finding the extreme value with respect to t gives the minimum value of I(t) as 21(1−ln2).
Answer:
I(t)=21e2t+t, the minimum value is 21(1−ln2).
18. (10 points) The bounded region Ω is enclosed by the plane 2x+y+2z=2 and the three coordinate planes, and Σ is taken with the outward orientation; compute I=∬Σ(2x+1)dydz+2ydzdx+3zdxdy.
Divergence Theorem
Approach:
This question tests the application of the divergence theorem. A surface integral of the second kind can be converted by the divergence theorem into a triple integral; the integrand is the divergence, which does not depend on the detailed shape of the surface, so only the volume of the region is needed.
Steps:
Identify the type of the integral: I=∬Σ(2x+1)dydz+2ydzdx+3zdxdy is a surface integral of the second kind with Σ oriented outwards, and by the divergence theorem it becomes ∭ΩdivAdV.
Compute the divergence: divA=∂x∂(2x+1)+∂y∂(2y)+∂z∂(3z)=2+2+3=7.
Compute the volume: Ω is enclosed by the plane 2x+y+2z=2 and the three coordinate planes, so it is a tetrahedron whose intercepts are 1,2,1, and V=61⋅1⋅2⋅1=31; hence I=7V=37.
Answer:
37.
19. (10 points) Given f(0)=1 and 0<f′(x)<21, consider the sequence xn+1=f(xn).
(I) Prove that ∑n=1∞∣xn+1−xn∣ converges absolutely;
(II) Prove that limn→∞xn exists and that 0<limn→∞xn<2.
Limit of a Sequence
Approach:
This question tests the convergence or divergence of a recursively defined sequence. In the first part the difference of two consecutive terms is related to half of the preceding difference, and the comparison with a geometric series shows absolute convergence; in the second part the limit is taken on the basis of that convergence, and the bound on f′ excludes both a zero limit and an overly large limit.
Steps:
By the mean value theorem, ∣xn+1−xn∣=∣f(xn)−f(xn−1)∣≤21∣xn−xn−1∣, hence ∣xn+1−xn∣≤(21)n−1∣x2−x1∣.
Comparing with the geometric series ∑(21)n−1, the comparison test shows that ∑n=1∞∣xn+1−xn∣ converges absolutely, which proves the first part.
By that convergence the sequence xn converges; let its limit be a. Taking limits gives a=f(a); moreover f(0)=1 and f is continuous, so a=0 does not satisfy a=f(a), and hence a>0.
From f(a)−f(0)=a−1 and the mean value theorem, a−1=f′(ξ)⋅a (0<ξ<a), and since f′(ξ)<21 we get a−1<2a, that is a<2, and hence 0<a<2.
Answer:
(I) ∑n=1∞∣xn+1−xn∣ converges absolutely; (II) limn→∞xn exists and 0<limn→∞xn<2.
20. (11 points) Let the matrices A=12−1−1a1−11a and B=21−a−12a−2. For which values of a does the equation AX=B have no solution, a unique solution, and infinitely many solutions? When it has solutions, solve the equation.
system of linear equationsrank of a matrixaugmented matrix
Approach:
This question tests the discussion of the solutions of the matrix equation AX=B. Treat B as the last column of the augmented matrix, reduce it by elementary row transformations, and decide the three cases by comparing r(A) with r(A:B).
Steps:
First compute the determinant of A:
∣A∣=12−1−1a1−11a=100−1a+20−13a−1=(a+2)(a−1)
so A is invertible when a=−2 and a=1.
Apply elementary row transformations to the augmented matrix:
(A:B)→100−1a+20−13a−12−31−a2a−40
When a=−2 and a=1, the third pivot a−1 is nonzero and the matrix reduces further to
10001000110−1a+23aa+2a−40
Then r(A)=r(A:B)=3, so the equation has a unique solution
X=A−1B=10−1a+23aa+2a−40
When a=1:
100−130−1302−302−30→1000100101−101−10
Then r(A)=r(A:B)=2<3, so the equation has infinitely many solutions, with general solution
X=1−k1−1k11−k2−1k2(k1,k2arbitrary constants)
When a=−2:
100−100−13−32−332−60→100−100−1102−10201
Then r(A)=2=r(A:B)=3, so the equation has no solution.
Answer:
For a=−2 and a=1 the equation has the unique solution X=10−1a+23aa+2a−40; for a=1 it has infinitely many solutions; for a=−2 it has no solution.
21. (11 points) Let A=020−1−30100.
(I) Find A99;
(II) Let B=(α1,α2,α3) be a 3rd-order matrix satisfying B2=BA. Write B100=(β1,β2,β3), and express β1,β2,β3 as linear combinations of α1,α2,α3.
eigenvalueeigenvectormatrix power
Approach:
This question tests matrix powers and the recursion of powers. The three eigenvalues of A are distinct, so A99 is obtained by diagonalisation; for the second part, B2=BA is iterated to give B100=BA99, and the columns are then expanded using the expression for A99.
Steps:
(I) The characteristic polynomial is
∣λE−A∣=λ−201λ+30−10λ=λ(λ+1)(λ+2)=0
so the eigenvalues are λ1=−1,λ2=−2,λ3=0.
For λ1=−1: −E−A=−1−20120−10−1, giving the eigenvector ξ1=110.
For λ2=−2: −2E−A=−2−20110−10−2, giving ξ2=120.
For λ3=0: −A=0−20130−100, giving ξ3=322.
Put P=(ξ1,ξ2,ξ3)=110120322; then
P−1AP=−1000−20000,P−1=2−10−110−22121
(II) Iterating B2=BA repeatedly:
B100=B98B2=B99A=⋯=BA99
Using part (I),
BA99=B299−22100−201−2991−210002−2982−2990=(β1β2β3)
Expanding each column of A99 in terms of the columns of B expresses β1,β2,β3 as linear combinations of α1,α2,α3.
Answer:
(I) A99=299−22100−201−2991−210002−2982−2990;
(II) from B100=BA99 together with part (I), expanding the three columns of A99 in terms of the columns of B gives β1,β2,β3 as linear combinations of α1,α2,α3.
22. (11 points) The two-dimensional random variable (X,Y) is uniformly distributed on D={(x,y)∣0<x<1,x2<y<x}, and U=1{XY≤1/4}.
(I) Write down the density of (X,Y); (II) decide whether U and X are independent and explain; (III) find the distribution function of Z=U+X.
Joint Distribution
Approach:
This question tests the density of a two-dimensional uniform distribution, the determination of the independence of random variables, and the piecewise computation of a distribution function. The three parts use, in order, the area method to get the density, a counterexample to decide independence by the definition of independence, and a case split according to the range of z and the value taken by U to find the distribution function.
Steps:
(I) The area of the region D={(x,y)∣0<x<1,x2<y<x} is ∫01(x−x2)dx=21, so the density of the uniform distribution on D is f(x,y)=∣D∣1=2, and it is 0 elsewhere.
(II) To test independence: U=1 corresponds to the event {XY≤41}, and the probability given by its area on D is not equal to the product of P(U=1) and P(X≤x), so U and X are not independent.
(III) To find FZ(z)=P(U+X≤z): split according to z<0, 0≤z<1, 1≤z<2 and z≥2, and within each piece split again according to the two possible values U=0 and U=1, integrating the distribution of X separately and adding.
Answer:
(I) f(x,y)={2,0,(x,y)∈Dother;
(II) not independent (the probability of the joint event is not equal to the product);
(III) write FZ(z) piecewise: it is 0 for z<0, for 0≤z<1 integrate separately over the two cases using X and U=0/1, the same holds for 1≤z<2, and it is 1 for z≥2.
23. (11 points) The population density is f(x;θ)=⎩⎨⎧θ33x2,0,0<x<θother, the sample is X1,X2,X3, and T=max{X1,X2,X3}.
(I) Find the density of T; (II) determine a so that aT is an unbiased estimator of θ.
Maximum Order Statistic
Approach:
This question tests the distribution function of an order statistic and the determination of the coefficient of an unbiased estimator. T is the maximum of the three sample values, so use the general result for order statistics FT(t)=FX(t)3; in the second part E(T) is then used to solve for the coefficient a.
Steps:
First find the distribution function of the population: FX(t)=∫0tθ33x2dx=θ3t3 (0<t<θ).
(I) The distribution function of T=max{X1,X2,X3} is FT(t)=FX(t)3=θ9t9, and differentiating gives the density fT(t)=θ99t8 (0<t<θ), and 0 elsewhere.
(II) Compute E(T)=∫0θt⋅θ99t8dt=θ99⋅11θ11=109θ; setting E(aT)=aE(T)=θ gives a=910, and then aT is an unbiased estimator of θ.
Answer:
(I) fT(t)=⎩⎨⎧θ99t8,0,0<t<θother;
(II) a=910.