This is a beta course, so its structure, chapters, and examples may continue to change.
2019 National Graduate Entrance Examination Mathematics I Past Paper
I. Multiple Choice
I. Multiple Choice: Questions 1-8, 4 points each, 32 points in total.
1. As x→0, if x−tanx and xk are infinitesimals of the same order, then k= ( )
(A) 1 (B) 2 (C) 3 (D) 4
order of infinitesimal
Approach:
This question tests the comparison of the orders of two infinitesimals by means of a Taylor expansion. Two infinitesimals are of the same order if and only if their ratio tends to a nonzero constant, so it is enough to expand tanx to a sufficiently high order.
Steps:
From the Maclaurin expansion tanx=x+3x3+o(x3) we get x−tanx=−3x3+o(x3).
Dividing by x3: x3x−tanx→−31=0, so x−tanx is of the same order as x3.
If it is of the same order as xk then k=3, so (C) is the correct choice.
Answer:
C.
2. Let the function f(x)={x∣x∣,xlnx,x≤0,x>0, then x=0 is ( ) of f(x)
(A) a point of differentiability, a point of extremum (B) a point of non-differentiability, a point of extremum (C) a point of differentiability, not a point of extremum (D) a point of non-differentiability, not a point of extremum
differentiabilityextremum
Approach:
This question tests the differentiability of a piecewise function at its breakpoint and the criterion for an extremum. The signs of the function values on the two sides of 0 (for the extremum) and the left and right derivatives (for the differentiability) have to be examined separately.
Steps:
f(0)=0. For x<0 we have f(x)=x∣x∣=−x2<0; for 0<x<1 we have f(x)=xlnx<0.
In a sufficiently small punctured neighbourhood of 0 we always have f(x)<0=f(0), so x=0 is a local maximum point.
The left derivative is x→0−limxf(x)−f(0)=x→0−lim(−x)=0.
The right derivative is x→0+limxf(x)−f(0)=x→0+limlnx=−∞, so no finite right derivative exists.
The one-sided derivatives are not equal, so f is not differentiable at x=0; combined with the fact that it is a local maximum point, this gives (B).
Answer:
B.
3. Let {un} be a monotonically increasing bounded sequence, then the series that converges is ( )
This question tests the convergence of series of numbers. The key is to recognise that option (D) is a telescoping series whose partial sums can be written down directly; the other options need not converge under the given assumptions.
Steps:
For option (D): its partial sums are SN=n=1∑N(un+12−un2)=uN+12−u12, since the intermediate terms cancel.
Since {un} is monotonically increasing and bounded, uN+12 has a limit, so SN converges, that is, the series n=1∑∞(un+12−un2) converges.
For option (A): take un=n, then the series becomes n=1∑∞nn=∑1, which diverges.
For option (B): take un=n, then the series becomes n=1∑∞(−1)nn1, which converges conditionally; while for un≡1 the same series converges, so (B) does not always converge.
For option (C): with un=n we have 1−un+1un=n+11, so the series diverges, and (C) does not always converge either. In conclusion the answer is (D).
Answer:
D.
4. Let the function Q(x,y)=y2x. If for every oriented smooth closed curve C in the upper half-plane (y>0) one has ∮CPdx+Qdy=0, then P(x,y) can be taken as ( )
(A) y−y3x2 (B) y1−y3x2 (C) x1−y1 (D) x−y1
conservative field
Approach:
This question tests the conditions under which a second-kind line integral in the plane is independent of the path. The condition that the integral vanishes for every closed curve is equivalent to the vector field being irrotational, that is, to Py=Qx holding identically.
Steps:
The upper half-plane y>0 is a simply connected region, so the necessary and sufficient condition for path independence is that ∂y∂P=∂x∂Q holds identically in that region.
From Q(x,y)=y2x we get Qx=y21, so we need Py=y21.
Integrating with respect to y: P(x,y)=∫y2dy+g(x)=−y1+g(x).
For options (A) and (B) the value of Py is not y21; for option (C) the function P=x1−y1 is not defined at x=0, so it fails the requirement of being defined on the whole upper half-plane.
In option (D), P=x−y1 corresponds to g(x)=x, which satisfies the condition, so (D) is the correct choice.
Answer:
D.
5. Let A be a real symmetric matrix of order 3 and E the identity matrix of order 3. If A2+A=2E and ∣A∣=4, then the canonical form of the quadratic form xTAx is ( )
This question tests how a matrix equation determines the eigenvalues and how the canonical form of a quadratic form is obtained. The relation A2+A=2E shows that every eigenvalue satisfies the same quadratic equation, and the determinant then fixes the spectrum.
Steps:
Let λ be an eigenvalue of A; from A2+A=2E we get λ2+λ−2=0, that is (λ−1)(λ+2)=0.
Hence the eigenvalues of A can only be 1 or −2.
A is a real symmetric matrix of order 3, so all eigenvalues are real and their multiplicities sum to 3; also ∣A∣=4>0, and with the eigenvalues 1 and −2 we get 1⋅(−2)2=4, which is consistent.
Hence the spectrum is {1,−2,−2}, the positive index of inertia is 1 and the negative index of inertia is 2, and the canonical form is y12−y22−y32, so (C) is the correct choice.
Answer:
C.
6. There are three planes intersecting pairwise, and their intersection lines are mutually parallel; let the equations ai1x+ai2y+ai3z=di(i=1,2,3) form a system of linear equations whose coefficient matrix and augmented matrix are denoted by A,A respectively, then ( )
This question tests the direction of the intersection lines of the planes and the rank of a system of linear equations. The fact that the three planes intersect pairwise and that their intersection lines are mutually parallel means that their normal vectors are linearly dependent, so the rank of the coefficient matrix is less than 3.
Steps:
The normal vectors of the three planes are (ai1,ai2,ai3). If the intersection lines are mutually parallel, then all three normal vectors are perpendicular to the same direction vector, so the three normal vectors are linearly dependent.
Hence the rank of the coefficient matrix satisfies r(A)≤2; and the three normal vectors cannot be collinear (otherwise the three planes would be parallel or coincident, which is a contradiction), so r(A)=2.
The rank r(A) of the augmented matrix depends on the right-hand side di: if r(A)=2, then the system has solutions and the solution space is a line (of dimension 1), that is, the three planes share a common line, in which case their intersection lines could not be mutually parallel, a contradiction.
Hence r(A)=3, that is, the system has no solution (the three planes have no common point). In conclusion r(A)=2,r(A)=3, so (A) is the correct choice.
Answer:
A.
7. Let A,B be random events, then the necessary and sufficient condition for P(A)=P(B) is ( )
This question tests the addition formula for probabilities and the relation between difference events. Rewrite P(A)=P(B) as an equality of the probabilities of two difference events and then compare with the options one by one.
Steps:
From P(A)=P(AB)+P(AB) and P(B)=P(AB)+P(AB), subtracting the two equations gives P(A)−P(B)=P(AB)−P(AB).
Hence P(A)=P(B) if and only if P(AB)=P(AB), which is exactly the content of option (C).
For option (A): P(A∪B)=P(A)+P(B) is equivalent to P(AB)=0, which only says that A,B are mutually exclusive, and that is not equivalent to P(A)=P(B).
For options (B) and (D): they correspond respectively to the independence of A,B and to a certain special relation, and neither of them implies P(A)=P(B). In conclusion the answer is (C).
Answer:
C.
8. Let the random variables X and Y be independent and both follow the normal distribution N(μ,σ2), then P{∣X−Y∣<1} ( )
(A) is independent of μ but depends on σ2 (B) depends on μ but not on σ2 (C) depends on both μ and σ2 (D) depends on neither μ nor σ2
normal distribution
Approach:
This question tests the linearity property of the normal distribution. The difference of two independent identically distributed normal variables is again normally distributed, and its mean and variance can be computed directly from the original distribution.
Steps:
Since X,Y are independent and both follow N(μ,σ2), we have X−Y∼N(μ−μ,σ2+σ2)=N(0,2σ2).
Hence P{∣X−Y∣<1}=P{2σX−Y<2σ1}.
The right-hand side contains σ but not μ, so this probability is independent of μ and depends on σ2, so (A) is the correct choice.
Answer:
A.
II. Fill in the Blanks
II. Fill in the Blanks: Questions 9-14, 4 points each, 24 points in total.
9. Let the function f(u) be differentiable, z=f(siny−sinx)+xy, then cosx1∂x∂z+cosy1∂y∂z=______.
partial derivative
Approach:
This question tests the computation and simplification of partial derivatives of a composite function. Here z is a function of u=siny−sinx, and after the two partial derivatives have been obtained by the chain rule the f′ terms cancel exactly.
Steps:
Put u=siny−sinx, then z=f(u)+xy, ux=−cosx, uy=cosy.
By the chain rule zx=f′(u)⋅(−cosx)+y and zy=f′(u)⋅cosy+x.
The f′(u) terms cancel each other, so the original expression =cosxy+cosyx.
Answer:
cosxy+cosyx.
10. The particular solution of the differential equation 2yy′−y2−2=0 satisfying y(0)=1 is y=______.
differential equation
Approach:
This question tests a differential equation with separable variables. Regard the left-hand side of the equation as a linear combination of the derivative of y2 and of y2 itself, and with the substitution u=y2 the equation becomes a first-order linear one.
Steps:
The original equation is 2yy′−y2−2=0, and we note that (y2)′=2yy′.
Put u=y2, then u′=2yy′, and the equation becomes u′−u−2=0, that is u′−u=2.
The homogeneous solution is u=Cex, and the particular solution is taken to be the constant u0=−2, so the general solution is u=Cex−2.
From y(0)=1 we get u(0)=1, and substitution gives 1=C−2, so C=3, that is u=3ex−2.
Since y(0)>0 we take the positive branch, giving y=3ex−2.
Answer:
3ex−2.
11. The sum function of the power series n=0∑∞(2n)!(−1)nxn on (0,+∞) is S(x)=______.
power series
Approach:
This question tests the sum function of a power series. The coefficients contain (2n)! while the variable appears as xn, which suggests the substitution t=x that turns the series into an even-power series.
Steps:
Put t=x (for x>0 we have t>0), then xn=t2n, and the original series becomes n=0∑∞(2n)!(−1)nt2n.
From the Taylor expansion of the cosine function cost=n=0∑∞(2n)!(−1)nt2n, the sum of the series is cost.
Substituting back t=x, the sum function is S(x)=cosx.
Answer:
cosx.
12. Let Σ be the upper side of the surface x2+y2+4z2=4(z≥0), then ∬Σ4−x2−4z2dxdy=______.
surface integral
Approach:
This question tests the direct computation of a second-kind surface integral on a surface. First simplify the integrand using the equation of the surface, then reduce the surface integral to a double integral on the xOy plane, and finally evaluate it in polar coordinates.
Steps:
From the surface equation x2+y2+4z2=4 we get 4−x2−4z2=y2, so the integrand is 4−x2−4z2=y2=∣y∣.
The projection of the upper half ellipsoid on the xOy plane is Dxy={(x,y)∣x2+y2≤4}. For the upper side (the positive orientation) the corresponding area element is dxdy, hence
∬Σ4−x2−4z2dxdy=∬Dxy∣y∣dxdy
Using polar coordinates x=rcosθ,y=rsinθ, 0≤r≤2, 0≤θ≤2π, the integrand becomes r∣sinθ∣ and the area element is rdrdθ, so
∬Dxy∣y∣dxdy=4∫02πdθ∫02r2sinθdr
13. Let A=(α1,α2,α3) be a matrix of order 3. If α1,α2 are linearly independent and α3=−α1+2α2, then the general solution of the system Ax=0 is ______.
system of linear equations
Approach:
This question tests the solution space of the system Ax=0. The solution vectors of Ax=0 are related to the linear combinations of the column vectors that vanish, so the given linear relation immediately yields the general solution.
Steps:
With A=(α1,α2,α3), the system Ax=0 is x1α1+x2α2+x3α3=0.
Since α1,α2 are linearly independent, x1−x3=0,x2+2x3=0, that is x1=x3,x2=−2x3.
Let x3=k∈R; the general solution is x=k(1,−2,1)T.
Answer:
k(1,−2,1)T,k∈R.
14. Let the random variable X have density f(x)={2x,0,0<x<2otherwise, let F(x) be the distribution function of X and EX the mathematical expectation of X, then P{F(X)>EX−1}=______.
distribution functionexpectation
Approach:
This question tests the properties of the distribution function and the computation of an expectation. First compute EX, then use the explicit expression of F(x) to rewrite the event {F(X)>EX−1} as an inequality in X and compute its probability.
Steps:
Compute the expectation:
EX=∫02x⋅2xdx=21∫02x2dx=21⋅38=34
so EX−1=31>0.
From f(x)=2x(0<x<2) the distribution function is, for 0≤x≤2,
F(x)=∫0x2tdt=4x2
while F(x)=0 for x<0 and F(x)=1 for x≥2.
Hence the event {F(X)>EX−1}={4X2>31}={X>32} (note that X>0).
III. Problems: Questions 15-23, 94 points in total.
15. (10 points) Let the function y(x) be the particular solution of the differential equation y′+xy=e−2x2 satisfying y(0)=0.
(1) Find y(x);
(2) Find the concavity intervals and the inflection points of the curve y=y(x).
first-order linear differential equationconcavity intervalsinflection point
Approach:
This question tests the solution of a first-order linear differential equation and the analysis of the concavity of the graph. First find the particular solution by the method of variation of constants (or the integrating factor), then compute the second derivative and use its sign to determine the concavity intervals and the inflection points.
Steps:
(1) The general solution of y′+xy=e−2x2 is
y=(∫e−2x2⋅e∫xdxdx+C)e−∫xdx=(∫e−2x2⋅e2x2dx+C)e−2x2=(x+C)e−2x2
From y(0)=0 we get C=0, so the required particular solution is y=xe−2x2.
(2) Differentiating gives y′=(1−x2)e−2x2.
Differentiating once more gives
y′′=(−2x)e−2x2+(1−x2)(−x)e−2x2=x(x2−3)e−2x2
Since e−2x2>0 always holds, the sign of y′′ is decided only by x(x2−3), whose zeros are x=−3,0,3.
From the sign changes: when x∈(−∞,−3) and (3,+∞) we have y′′>0, so the curve is convex; when x∈(−3,0) and (0,3) we have y′′<0, so the curve is concave.
Since y′′ changes sign at each of x=−3,0,3, and the curve is continuous at these points, each of them is an inflection point.
Answer:
(1) y=xe−2x2;
(2) The convex intervals are (−∞,−3) and (3,+∞), and the concave intervals are (−3,0) and (0,3); the inflection points are (−3,3e−23), (0,0) and (3,3e−23).
16. (10 points) Let a,b be real numbers. For the function z=2+ax2+by2, among the directional derivatives at the point (3,4), the one along the direction l=−3i−4j is the largest, and its maximum value is 10.
(1) Find a,b;
(2) Find the area of the surface z=2+ax2+by2(z≥0).
directional derivativegradientsurface area
Approach:
This question tests the relation between directional derivatives and the gradient, as well as the computation of a surface area. The directional derivative attains its maximum in the direction of the gradient, so first use the maximum direction together with the maximum value to determine a,b, and then turn the surface into a paraboloid of revolution and compute its area in polar coordinates.
Steps:
(1) gradz=(2ax,2by), so at the point (3,4) we have gradz=(6a,8b).
The directional derivative attains its maximum in the direction of the gradient, and that maximum equals ∣gradz∣. Since the largest directional derivative is along l=(−3,−4) and equals 10, the gradient must point in the same direction as l, hence
−36a=−48b
that is a=b.
Also ∣gradz∣=36a2+64b2=10; substituting a=b gives 10∣a∣=10, that is a=±1. Since the value taken is the maximum10>0, the gradient must have the same direction as l, i.e. 6a<0, so a=b=−1.
(2) The surface is then z=2−x2−y2(z≥0), whose projection on the xOy plane is Dxy={(x,y)∣x2+y2≤2}.
The surface element is 1+zx2+zy2dxdy=1+4x2+4y2dxdy, so
S=∬Dxy1+4x2+4y2dxdy
Using polar coordinates x=rcosθ,y=rsinθ, 0≤r≤2, 0≤θ≤2π, we get
S=2π∫02r1+4r2dr=2π⋅121(1+4r2)2302=6π(93−1)=32π(33−1)
Answer:
(1) a=b=−1;
(2) S=32π(33−1).
17. (10 points) Find the area of the region bounded by the curve y=e−xsinx(x≥0) and the x-axis.
improper integralarea
Approach:
This question tests improper integrals and the integration of a periodic function over an unbounded interval. Since ∣sinx∣ is periodic with period π, split the integral into intervals of length π and sum the resulting series.
Steps:
Since e−x>0 for x≥0 and ∣sinx∣≥0, the curve always lies on or above the x-axis, and the required area is the improper integral
A=∫0+∞e−x∣sinx∣dx
On [kπ,(k+1)π] we have ∣sinx∣=(−1)ksinx, so
A=limn→∞∑k=0n(−1)k∫kπ(k+1)πe−xsinxdx
Since ∫e−xsinxdx=−21e−x(sinx+cosx)+C,
∫kπ(k+1)πe−xsinxdx=[−21e−x(sinx+cosx)]kπ(k+1)π=21[e−kπ(−1)k−e−(k+1)π(−1)k+1]
Substituting and using [(−1)k]2=1, we get
A=21limn→∞∑k=0n[e−(k+1)π+e−kπ]=21limn→∞[1+2∑k=1ne−kπ+e−(n+1)π]
Since e−(n+1)π→0 and by the geometric-series formula k=1∑∞e−kπ=1−e−πe−π,
A=21(1+1−e−π2e−π)=21(1+eπ−12)=21+eπ−11
Answer:
21+eπ−11.
18. (10 points) Let an=∫01xn1−x2dx(n=0,1,2,⋯).
(1) Prove that the sequence {an} is monotonically decreasing and that an=n+2n−1an−2(n=2,3,⋯);
(2) Find n→∞liman−1an.
definite integralsequence limitsqueeze theorem
Approach:
This question tests the properties of a parameter-dependent definite integral and the computation of a sequence limit. In the first part monotonicity is obtained by comparing adjacent terms in the integral representation, and the recurrence is derived by integration by parts; in the second part the limit is squeezed out of the recurrence.
Steps:
(1) Monotonicity: for 0<x<1 we have xn>xn+1 and 1−x2>0, hence
an−an+1=∫01(xn−xn+1)1−x2dx=∫01xn(1−x)1−x2dx>0
so an>an+1, i.e. {an} is monotonically decreasing.
The recurrence: integrate by parts with u=xn and dv=x1−x2dx, so dv=−31d(1−x2)23; for n≥2,
an=∫01xn−1⋅x1−x2dx=−31∫01xn−1d(1−x2)23=3n−1∫01xn−2(1−x2)23dx
Also note that
an−2=∫01xn−2(1−x2)21dx=∫01xn−2(1−x2)23⋅1−x21dx
hence
n+2n−1an−2=n+2n−1∫01xn−2(1−x2)23⋅1−x21dx
Comparing the two expressions and using (1−x2)=(1−x)(1+x) gives an=n+2n−1an−2.
(2) By the monotonic decrease in (1) we have an<an−1, that is an−1an<1; and from the recurrence an−2an=n+2n−1, so
n+2n−1an−2<an−1an⟹n+2n−1⋅an−1an−2<an−1an<1
Since {an} is monotonically decreasing and bounded, an−1an−2≤1, and therefore
n+2n−1<an−1an<1
Taking limits on both sides, the squeeze theorem gives n→∞liman−1an=1.
Answer:
(1) {an} is monotonically decreasing, and an=n+2n−1an−2(n≥2);
(2) n→∞liman−1an=1.
19. (10 points) Let Ω be the cone enclosed by the conical surface x2+(y−z)2=(1−z)2(0≤z≤1) and the plane z=0. Find the coordinates of the centroid of Ω.
triple integralcentroid
Approach:
This question tests computing the centroid of a solid with a triple integral. The key is to slice the cone by the planes z=const: the conical surface equation shows that the cross-section is a disc with centre (0,z) and radius 1−z, and this reduces the triple integral to a one-variable integral.
Steps:
From the conical surface equation x2+(y−z)2=(1−z)2, when z is fixed the cross-section is the disc with centre (0,z) and radius r(z)=1−z. Hence Ω is symmetric about the yz-plane and xˉ=0.
The volume of the cone is
V=∭Ωdxdydz=∫01dz∬x2+(y−z)2≤(1−z)2dxdy=π∫01(1−z)2dz=3π
Compute ∭Ωydxdydz. On each cross-section write y=z+u, where u is symmetric about the centre of the disc and ∬udxdy=0, so
∬sectionydxdy=z⋅π(1−z)2
and therefore
∭Ωydxdydz=π∫01z(1−z)2dz=12π
By the centroid formulas xˉ=V1∭ΩxdV, yˉ=V1∭ΩydV, zˉ=V1∭ΩzdV,
xˉ=0,yˉ=π/3π/12=41,zˉ=π/3π/12=41
Answer:
The centroid is (0,41,41).
20. (11 points) Let the vectors α1=(1,2,1)T,α2=(1,3,2)T,α3=(1,a,3)T form a basis of R3, and let β=(1,1,1)T have coordinates (b,c,1)T in this basis.
(1) Find a,b,c;
(2) Prove that α2,α3,β form a basis of R3, and find the transition matrix from α2,α3,β to α1,α2,α3.
vector systembasistransition matrix
Approach:
This question tests the linear representation of vectors, the criterion for a basis, and transition matrices. In the first part the equation β=bα1+cα2+α3 gives a system of linear equations to solve; in the second part a determinant decides whether the three vectors form a basis, and the transition matrix is obtained by writing the new basis vectors in the old basis.
Steps:
(1) Since β=(1,1,1)T has coordinates (b,c,1)T in the basis (α1,α2,α3),
β=bα1+cα2+α3
that is
⎩⎨⎧b+c+1=1,2b+3c+a=1,b+2c+3=1.
The first equation gives b+c=0 and the third gives b+2c=−2; subtracting yields c=−2,b=2, and substituting into the second gives a=3.
(2) Compute the determinant
∣α2,α3,β∣=132133111=1(3−3)−1(3−2)+1(9−6)=2=0
so α2,α3,β are linearly independent, hence form a basis of R3.
To find the transition matrix, let the old and the new basis matrices be O=(α1,α2,α3) and N=(α2,α3,β):
O=121132133,N=132133111
Computing the inverse gives
N−1=0−21231−21−21−110
Hence the transition matrix from α2,α3,β to α1,α2,α3 is
Q=N−1O=1−2121100010
that is (α1,α2,α3)=(α2,α3,β)Q.
Answer:
(1) a=3,b=2,c=−2;
(2) α2,α3,β form a basis of R3, and the transition matrix is 1−2121100010.
21. (11 points) Suppose the matrices A=−220−2x01−2−2 and B=2001−1000y are similar.
(1) Find x,y;
(2) Find an invertible matrix P such that P−1AP=B.
similar matriceseigenvalueeigenvector
Approach:
This question tests the necessary conditions for similar matrices and the construction of a similarity transformation. In the first part the equal traces and the equal determinants of similar matrices give a system of equations for x,y; in the second part the eigenvectors of A and of B are computed and combined by the change-of-basis formula.
Steps:
(1) Similar matrices have the same eigenvalues, hence the same trace: −2+x−2=2−1+y, i.e. x−4=y+1, so y=x−5.
Similar matrices also have the same determinant: ∣A∣=4x−8 and ∣B∣=−2y, so 4x−8=−2y, i.e. y=−2x+4.
Solving y=x−5 together with y=−2x+4 gives x=3,y=−2.
(2) Then A=−220−2301−2−2 and B=2001−1000−2, and both have eigenvalues 2,−1,−2 (since ∣λE−A∣=(λ−2)(λ+1)(λ+2)).
For A, from 2E+A→020−210100 we get the eigenvector ξ1=(−1,2,4)T for λ=2; from E+A→100200−110 we get ξ2=(−2,1,0)T for λ=−1; and from 2E−A we get ξ3=(−1,2,0)T for λ=−2.
For B, from 2E+B→400110000 we get η1=(0,0,1)T for λ=2; from E+B we get η2=(−1,3,0)T for λ=−1; and from 2E−B we get η3=(1,0,0)T for λ=−2.
Take P=(ξ1,ξ2,ξ3)(η1,η2,η3)−1; then P−1AP=B. Here
(ξ1,ξ2,ξ3)=−124−210−120,(η1,η2,η3)=001−130100
and
P=−124−210−120001−130100−1=−120−110−124
One checks that P−1AP=2001−1000−2=B, and ∣P∣=4=0, so P is invertible.
Answer:
(1) x=3,y=−2;
(2) One may take P=−120−110−124, for which P−1AP=B.
22. (11 points) Let the random variables X and Y be independent, X∼Exp(1), P{Y=−1}=p,P{Y=1}=1−p(0<p<1), and let Z=XY.
(1) Find the density of Z;
(2) For which value of p are X and Z uncorrelated;
(3) Are X and Z independent? Explain.
random variablecorrelationindependence
Approach:
This question tests the density, the covariance and the independence of functions of random variables. The sign of Z=XY is completely decided by the value of Y, so X has to be transformed separately for the two values of Y, and the density is then obtained with the law of total probability; correlation uses the covariance formula, and independence is verified by an event-probability argument.
Steps:
(1) When Y=1 (probability 1−p) we have Z=X>0, whose contribution to the density is (1−p)e−z; when Y=−1 (probability p) we have Z=−X<0, and then fZ(z)=pfX(−z)=pez. Hence
fZ(z)={pez,(1−p)e−z,z<0,z>0.
(2) Since X∼Exp(1) we have E(X)=1,E(X2)=2; also E(Y)=(−1)p+1⋅(1−p)=1−2p.
As Z=XY and X,Y are independent, E(XZ)=E(X2)E(Y)=2(1−2p) and E(Z)=E(X)E(Y)=1−2p.
Hence Cov(X,Z)=E(XZ)−E(X)E(Z)=2(1−2p)−(1−2p)=1−2p, and setting this equal to zero gives p=21.
(3) X and Z are not independent. Note that X>0 always holds; if X≤1 and Z=XY≤−1, then necessarily Y=−1 and X≥1, contradicting X≤1 (since X is continuous, P{X=1}=0), so
P{X≤1,Z≤−1}=P{X≤1,XY≤−1}=0
On the other hand P{X≤1}>0 and P{Z≤−1}=P{Y=−1}P{X≥1}=p⋅e−1>0, so
P{X≤1,Z≤−1}=0=P{X≤1}⋅P{Z≤−1}
and therefore X and Z are not independent.
Answer:
(1) fZ(z)={pez,(1−p)e−z,z<0z>0;
(2) X and Z are uncorrelated when p=21;
(3) X and Z are not independent, because P{X≤1,Z≤−1}=0 while P{X≤1}P{Z≤−1}>0.
23. (11 points) Let the population X have density
f(x;σ2)=⎩⎨⎧σAe−2σ2(x−μ)2,0,x⩾μ,x<μ,
where μ is a known parameter, σ>0 is an unknown parameter and A is a constant. X1,X2,⋯,Xn is a simple random sample from the population X.
(1) Find A;
(2) Find the maximum likelihood estimator of σ2.
probability densitymaximum likelihood estimator
Approach:
This question tests the normalization of a probability density and the maximum likelihood estimation. In the first part the normalizing condition of the density is used to determine the constant A; in the second part the likelihood function is written, its logarithm is taken, and the unknown parameter is found by setting the derivative equal to zero.
Steps:
(1) By the normalizing condition,
1=∫μ+∞σAe−2σ2(x−μ)2dx=A∫μ+∞e−21(σx−μ)2d(σx−μ)=A∫0+∞e−2u2du
where we used ∫0+∞e−2u2du=22π (the even function and the normalizing property of the normal density). Hence
1=22πA,A=π2
(2) Since X1,…,Xn is a simple random sample, the likelihood function is
L(σ2)=∏i=1nσAe−2σ2(xi−μ)2=(σ2)2nAne−2σ21∑i=1n(xi−μ)2
Differentiating with respect to σ2 and setting the result equal to zero:
dσ2dlnL(σ2)=−2n⋅σ21+2σ41∑i=1n(xi−μ)2=0
Solving, the maximum likelihood estimate of σ2 is
σ^2=n1∑i=1n(xi−μ)2
Moreover the second derivative d(σ2)2d2lnL=2σ4n−σ61i=1∑n(xi−μ)2 is positive at this point, so the stationary point is indeed a maximum. Hence the maximum likelihood estimator of σ2 is
σ˙2=n1∑i=1n(Xi−μ)2
Answer:
(1) A=π2;
(2) The maximum likelihood estimator of σ2 is n1i=1∑n(Xi−μ)2.