2017 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 function f(x)={1−cos⁡xax,x>0b,x≤0f(x)=\begin{cases}\dfrac{1-\cos\sqrt{x}}{ax},&x>0\\[4pt] b,&x\le0\end{cases} is continuous at x=0x=0, then ( )

(A) ab=12ab=\dfrac{1}{2}
(B) ab=−12ab=-\dfrac{1}{2}
(C) ab=0ab=0
(D) ab=2ab=2

Continuity

Approach:

This question tests the determination of the continuity of a piecewise function at its breakpoint. The one-sided limits must both exist and must equal the value that the function takes at that point according to the corresponding piece; the key in this question is to compute the right-hand limit first with equivalent infinitesimals.

Steps:

  1. The breakpoint is x=0x=0. When x≤0x\le 0 we have f(x)=bf(x)=b, so f(0)=bf(0)=b.
  2. Compute the right-hand limit: lim⁡x→0+1−cos⁡xax\lim\limits_{x\to0^+}\dfrac{1-\cos\sqrt{x}}{ax}. Put t=x→0+t=\sqrt{x}\to0^+; the equivalent infinitesimal 1−cos⁡t∼t221-\cos t\sim\tfrac{t^2}{2} gives 1−cos⁡x∼x21-\cos\sqrt{x}\sim\tfrac{x}{2}, so lim⁡x→0+1−cos⁡xax=lim⁡x→0+12a=12a\lim\limits_{x\to0^+}\dfrac{1-\cos\sqrt{x}}{ax}=\lim\limits_{x\to0^+}\dfrac{1}{2a}=\dfrac{1}{2a}.
  3. The condition of continuity is that the right-hand limit equals f(0)=bf(0)=b, that is b=12ab=\dfrac{1}{2a}, and hence ab=12ab=\tfrac12.

Answer:

A.

2. Suppose the function f(x)f(x) is differentiable and f(x)f′(x)>0f(x)f'(x)>0; then ( )

(A) f(1)>f(−1)f(1)>f(-1)
(B) f(1)<f(−1)f(1)<f(-1)
(C) ∣f(1)∣>∣f(−1)∣|f(1)|>|f(-1)|
(D) ∣f(1)∣<∣f(−1)∣|f(1)|<|f(-1)|

Monotonicity

Approach:

This question tests the determination of the monotonicity of a function from the sign of its derivative. From f(x)f′(x)>0f(x)f'(x)>0 the monotonicity of ff cannot be read off directly, but after squaring ff the derivative of the new function is exactly 2f(x)f′(x)>02f(x)f'(x)>0, and the problem is then solved.

Steps:

  1. Put g(x)=f2(x)g(x)=f^2(x); then g′(x)=2f(x)f′(x)>0g'(x)=2f(x)f'(x)>0 holds identically, so gg is strictly increasing on the whole real axis.
  2. From 1>−11>-1 and the monotonicity of gg we get g(1)>g(−1)g(1)>g(-1), that is f2(1)>f2(−1)f^2(1)>f^2(-1).
  3. Taking square roots on both sides gives ∣f(1)∣>∣f(−1)∣|f(1)|>|f(-1)|, so option (C) is correct.

Answer:

C.

3. The directional derivative of the function f(x,y,z)=x2y+z2f(x,y,z)=x^2y+z^2 at the point (1,2,0)(1,2,0) along the direction of the vector n=(1,2,2)n=(1,2,2) is ( )

(A) 12 (B) 6 (C) 4 (D) 2

Directional Derivative

Approach:

This question tests the directional derivative of a function of several variables along a prescribed direction. The directional derivative is the inner product of the gradient with the unit vector in that direction, so it suffices to compute the gradient first and then to normalize the direction vector.

Steps:

  1. Compute the partial derivatives: fx=2xy, fy=x2+2zf_x=2xy,\ f_y=x^2+2z, so ∇f(1,2,0)=(4,1,0)\nabla f(1,2,0)=(4,1,0).
  2. The length of the direction vector n=(1,2,2)n=(1,2,2) is 1+4+4=3\sqrt{1+4+4}=3, and the corresponding unit vector is e=(13,23,23)\bm{e}=\left(\tfrac13,\tfrac23,\tfrac23\right).
  3. The directional derivative Def=∇f⋅e=4⋅13+1⋅23+0⋅23=2\bm{D}_{\bm{e}}f=\nabla f\cdot\bm{e}=4\cdot\tfrac13+1\cdot\tfrac23+0\cdot\tfrac23=2, so option (D) is correct.

Answer:

D.

4. In a race between A and B, A leads by 10 m when the timing starts; the solid line is the velocity curve v1(t)v_1(t) of A and the dashed line is the velocity curve v2(t)v_2(t) of B, and the three shaded areas are in turn 10, 20, 3. The moment t0t_0 at which B catches up with A lies in ( )

(A) (0,10)(0,10) (B) (10,25)(10,25) (C) (25,30)(25,30) (D) t0=25t_0=25

Application of the Definite Integral

Approach:

This question tests the use of a definite integral to represent the difference of the displacements and the determination of the moment of catching up. The area under a velocity curve is the displacement, and the condition of catching up is that the distance accumulated by B exceeds that of A by the 10 m lead held at the start of the timing.

Steps:

  1. The displacements of A and B at the moment tt are ∫0tv1(τ)dτ\int_0^t v_1(\tau)d\tau and ∫0tv2(τ)dτ\int_0^t v_2(\tau)d\tau respectively; since A leads by 10 m when the timing starts, B catching up with A is equivalent to the distance accumulated by B exceeding that of A by 10 m.
  2. Write S(t)=∫0t(v1(τ)−v2(τ))dτS(t)=\int_0^t\bigl(v_1(\tau)-v_2(\tau)\bigr)d\tau; then the condition of catching up is S(t)=−10S(t)=-10, and the increment of SS is exactly the oriented area between the two velocity curves, namely the shaded part of the figure.
  3. From the figure, on (0,10)(0,10) A is faster than B (shaded area 10), on (10,25)(10,25) B is faster than A (shaded area 20), and on (25,30)(25,30) A is faster than B again (shaded area 3), so SS first increases, then decreases, and then increases again.
  4. During the first 25 s the distance that B gains by its advantage in speed already exceeds 10 m, and S(t)S(t) inside (10,25)(10,25) crosses −10-10 from above; after 2525 s the relative position of the two curves has already changed, so the moment of catching up t0t_0 lies strictly inside (10,25)(10,25), and option (B) is correct.

Answer:

B.

5. Let α\alpha be a unit column vector in nn dimensions and EE the nnth order identity matrix; then ( )

(A) E−ααTE-\alpha\alpha^T is not invertible
(B) E+ααTE+\alpha\alpha^T is not invertible
(C) E−2ααTE-2\alpha\alpha^T is not invertible
(D) E+2ααTE+2\alpha\alpha^T is not invertible

Matrix

Approach:

This question tests the determination of the singularity of matrices containing a unit column vector. The rank of ααT\alpha\alpha^T is 1, so one can construct a nonzero vector belonging to the null space of the matrix and thereby decide directly that the corresponding matrix is not invertible.

Steps:

  1. Since α\alpha is a nonzero unit column vector, consider the matrix E−ααTE-\alpha\alpha^T acting on α\alpha.
  2. From αTα=1\alpha^T\alpha=1 we get (E−ααT)α=α−α(αTα)=α−α=0(E-\alpha\alpha^T)\alpha=\alpha-\alpha(\alpha^T\alpha)=\alpha-\alpha=0.
  3. Since α≠0\alpha\ne0, the homogeneous equation (E−ααT)x=0(E-\alpha\alpha^T)x=0 has a nonzero solution, so the determinant of the matrix is 0 and the matrix is not invertible; option (A) is correct.

Answer:

A.

6. Let A=(200021001)A=\begin{pmatrix}2&0&0\\0&2&1\\0&0&1\end{pmatrix}, B=(210020001)B=\begin{pmatrix}2&1&0\\0&2&0\\0&0&1\end{pmatrix}, C=(100020002)C=\begin{pmatrix}1&0&0\\0&2&0\\0&0&2\end{pmatrix}; then ( )

(A) A and C are similar, B and C are similar
(B) A and C are similar, B and C are not similar
(C) A and C are not similar, B and C are similar
(D) A and C are not similar, B and C are not similar

Matrix Similarity

Approach:

This question tests the determination of matrix similarity and of diagonalization. Two matrices are similar only if they have the same eigenvalues (with multiplicities) and each of them is diagonalizable; the latter is equivalent to saying that for every eigenvalue the dimension of the corresponding eigenspace equals its multiplicity.

Steps:

  1. The eigenvalues of A=(200021001)A=\begin{pmatrix}2&0&0\\0&2&1\\0&0&1\end{pmatrix} and of C=(100020002)C=\begin{pmatrix}1&0&0\\0&2&0\\0&0&2\end{pmatrix} are both 2,2,12,2,1, so AA and CC are similar.
  2. The eigenvalues of B=(210020001)B=\begin{pmatrix}2&1&0\\0&2&0\\0&0&1\end{pmatrix} are also 2,2,12,2,1, but the eigenspace of the eigenvalue 22 has dimension 1, which is smaller than the multiplicity 2, so BB is not diagonalizable.
  3. A diagonalizable matrix is necessarily similar to its diagonal matrix; since BB is not diagonalizable, BB and CC are not similar, and so option (B) is correct.

Answer:

B.

7. Let A,BA,B be random events with P(A)>0,P(B)>0P(A)>0,P(B)>0; then the necessary and sufficient condition for P(A∣B)=P(A)P(A\mid B)=P(A) is ( )

(A) P(B∣A)=P(B)P(B\mid A)=P(B)
(B) P(A∣B‾)=P(A)P(A\mid \overline{B})=P(A)
(C) P(B∣A‾)=P(B)P(B\mid \overline{A})=P(B)
(D) P(A∣B‾)=P(A‾)P(A\mid \overline{B})=P(\overline{A})

Independence

Approach:

This question tests the definition of the independence of events and its equivalent conditions. P(A∣B)=P(A)P(A\mid B)=P(A) is precisely the equivalent description of the independence of A,BA,B, so it suffices to compare this condition with the condition in each of the options.

Steps:

  1. From the formula for the conditional probability P(A∣B)=P(AB)P(B)P(A\mid B)=\dfrac{P(AB)}{P(B)} (P(B)>0P(B)>0), we get P(A∣B)=P(A)  ⟺  P(AB)=P(A)P(B)P(A\mid B)=P(A)\iff P(AB)=P(A)P(B), that is AA and BB are independent.
  2. Under the condition P(A)>0P(A)>0, the independence of A,BA,B is in turn equivalent to P(B∣A)=P(AB)P(A)=P(B)P(B\mid A)=\dfrac{P(AB)}{P(A)}=P(B), so option (A) is equivalent to the condition given in the question.
  3. Options (B), (C) and (D) describe the independence of A,BA,B and of the complementary events, and they cannot imply P(A∣B)=P(A)P(A\mid B)=P(A); ruling them out one by one, the answer is (A).

Answer:

A.

8. Let X1,…,Xn (n≥2)X_1,\dots,X_n\ (n\ge2) be a sample from N(μ,σ2)N(\mu,\sigma^2) and let X‾=1n∑Xi\overline{X}=\frac{1}{n}\sum X_i; which of the following conclusions is incorrect ( )

(A) ∑i=1n(Xi−μ)2σ2∼χ2(n)\sum\limits_{i=1}^n\dfrac{(X_i-\mu)^2}{\sigma^2}\sim\chi^2(n)
(B) ∑i=1n(Xi−X‾)2σ2∼χ2(n)\sum\limits_{i=1}^n\dfrac{(X_i-\overline{X})^2}{\sigma^2}\sim\chi^2(n)
(C) (n−1)S2σ2∼χ2(n−1)\dfrac{(n-1)S^2}{\sigma^2}\sim\chi^2(n-1)
(D) n(X‾−μ)S∼t(n−1)\dfrac{\sqrt{n}(\overline{X}-\mu)}{S}\sim t(n-1)

Sampling Distribution

Approach:

This question tests the degrees of freedom of the usual sampling distributions for a normal population. The key is to remember that the sum of the squared deviations from the sample mean divided by σ2\sigma^2 has n−1n-1 degrees of freedom and not nn; the other three conclusions are standard results.

Steps:

  1. When the XiX_i come from N(μ,σ2)N(\mu,\sigma^2) and μ\mu is known, ∑i=1n(Xi−μ)2σ2∼χ2(n)\sum\limits_{i=1}^n\dfrac{(X_i-\mu)^2}{\sigma^2}\sim\chi^2(n), so option (A) is correct.
  2. Replacing the population mean by the sample mean loses one degree of freedom, and ∑i=1n(Xi−X‾)2σ2∼χ2(n−1)\sum\limits_{i=1}^n\dfrac{(X_i-\overline{X})^2}{\sigma^2}\sim\chi^2(n-1), so the degrees of freedom nn in option (B) is incorrect.
  3. The sample variance S2=1n−1∑i=1n(Xi−X‾)2S^2=\dfrac{1}{n-1}\sum\limits_{i=1}^n(X_i-\overline{X})^2 satisfies (n−1)S2σ2∼χ2(n−1)\dfrac{(n-1)S^2}{\sigma^2}\sim\chi^2(n-1), and T=n(X‾−μ)S∼t(n−1)T=\dfrac{\sqrt{n}(\overline{X}-\mu)}{S}\sim t(n-1), so (C) and (D) are correct.

Answer:

B.

Part II: Fill in the Blank

Part II: Fill in the Blank: Questions 9-14, 4 points each, 24 points in total.

9. Given the function f(x)=11+x2f(x)=\dfrac{1}{1+x^2}, then f(3)(0)=f^{(3)}(0)= ______.

Higher-Order Derivative

Approach:

This question tests the use of a power series expansion to compute the value of a higher-order derivative at a point. The expansion of f(x)=11+x2f(x)=\dfrac{1}{1+x^2} contains only even powers, and the coefficient of the x3x^3 term is f(3)(0)f^{(3)}(0) divided by 3!3!.

Steps:

  1. Expand f(x)f(x) into a power series: 11+x2=1−x2+x4−x6+⋯\dfrac{1}{1+x^2}=1-x^2+x^4-x^6+\cdots, which is the geometric series ∑n=0∞(−1)nx2n\sum\limits_{n=0}^{\infty}(-1)^nx^{2n}.
  2. By the relation between a power series and a Taylor series, the coefficient of the x2nx^{2n} term is f(2n)(0)(2n)!\dfrac{f^{(2n)}(0)}{(2n)!}, and here only even powers occur.
  3. The coefficient of the x3x^3 term in the expansion is 0, so f(3)(0)3!=0\dfrac{f^{(3)}(0)}{3!}=0, that is f(3)(0)=0f^{(3)}(0)=0.

Answer:

10. The general solution of the differential equation y′′+2y′+5y=0y''+2y'+5y=0 is ______.

Ordinary Differential Equation

Approach:

This question tests the computation of the general solution of a second-order linear differential equation with constant coefficients. First write the characteristic equation and find its roots, and then apply the corresponding form of the general solution according to the type of the roots.

Steps:

  1. The characteristic equation is λ2+2λ+5=0\lambda^2+2\lambda+5=0, and its discriminant is Δ=4−20=−16<0\Delta=4-20=-16<0, so it has a pair of conjugate complex roots.
  2. By the quadratic formula λ1,2=−2±4i2=−1±2i\lambda_{1,2}=\dfrac{-2\pm4i}{2}=-1\pm2i, that is α=−1, β=2\alpha=-1,\ \beta=2.
  3. The general solution containing a pair of conjugate complex roots α±βi\alpha\pm\beta i is y=eαx(C1cos⁡βx+C2sin⁡βx)y=e^{\alpha x}\bigl(C_1\cos\beta x+C_2\sin\beta x\bigr), and substituting gives y=e−x(C1cos⁡2x+C2sin⁡2x)y=e^{-x}(C_1\cos2x+C_2\sin2x).

Answer:

y=e−x(C1cos⁡2x+C2sin⁡2x)y=e^{-x}(C_1\cos2x+C_2\sin2x).

11. If the curve integral ∫Lx dy−ay dxx2+y2\displaystyle \int_L \dfrac{x\,dy-a y\,dx}{x^2+y^2} is path independent in the region D={(x,y)∣x2+y2>0}D=\{(x,y)\mid x^2+y^2>0\}, then a=a= ______.

Path Independence

Approach:

This question tests the condition for a curve integral of the second kind to be path independent. Path independence is equivalent to Py=QxP_y=Q_x holding identically in the simply connected region corresponding to the integrand, and from this the parameter aa is solved.

Steps:

  1. Write the integral in the form P dx+Q dyP\,dx+Q\,dy: P(x,y)=−ayx2+y2P(x,y)=-\dfrac{ay}{x^2+y^2} and Q(x,y)=xx2+y2Q(x,y)=\dfrac{x}{x^2+y^2}.
  2. The region D={(x,y)∣x2+y2>0}D=\{(x,y)\mid x^2+y^2>0\} is simply connected, so the necessary and sufficient condition for the integral to be path independent is that Py=QxP_y=Q_x holds identically in DD.
  3. Compute the two partial derivatives above and set them equal: the denominators of the two expressions are both (x2+y2)2(x^2+y^2)^2, and after cancelling the common factor one obtains a linear equation in aa.
  4. Solving this equation gives a=−1a=-1, and substituting it back verifies the path-independence condition, so the blank should be filled with −1-1.

Answer:

−1-1.

12. The sum function of the power series ∑n=0∞(−1)nx2n+1(2n+1)(2n+1)!\displaystyle \sum_{n=0}^{\infty}\frac{(-1)^n x^{2n+1}}{(2n+1)(2n+1)!} is ______.

Power Series

Approach:

This question tests the computation of the sum function of a power series. The given coefficients contain (2n+1)(2n+1)!(2n+1)(2n+1)!, and differentiating term by term turns the series into the familiar sine series, after which the constant of integration is fixed by the initial value.

Steps:

  1. Write S(x)=∑n=0∞(−1)nx2n+1(2n+1)(2n+1)!S(x)=\sum\limits_{n=0}^{\infty}\dfrac{(-1)^nx^{2n+1}}{(2n+1)(2n+1)!}; differentiating term by term inside the radius of convergence gives S′(x)=∑n=0∞(−1)nx2n(2n+1)!S'(x)=\sum\limits_{n=0}^{\infty}\dfrac{(-1)^nx^{2n}}{(2n+1)!}.
  2. From the expansion of the sine function sin⁡x=∑n=0∞(−1)nx2n+1(2n+1)!\sin x=\sum\limits_{n=0}^{\infty}\dfrac{(-1)^nx^{2n+1}}{(2n+1)!} we know that S′(x)=sin⁡xxS'(x)=\dfrac{\sin x}{x} (x≠0x\ne0).
  3. Integrating S′(x)S'(x) and using the initial value S(0)=0S(0)=0 to fix the constant of integration, the sum function obtained is sin⁡xx\dfrac{\sin x}{x}.
  4. When x=0x=0 the value 1 is taken, as follows from lim⁡x→0sin⁡xx=1\lim\limits_{x\to0}\dfrac{\sin x}{x}=1, so the sum function is sin⁡xx\dfrac{\sin x}{x} (x≠0x\ne0, and it is 1 when x=0x=0).

Answer:

sin⁡xx\displaystyle \frac{\sin x}{x} (x≠0x\ne0, and it is 1 when x=0x=0).

13. Let A=(101112011)\boldsymbol{A}=\begin{pmatrix}1&0&1\\1&1&2\\0&1&1\end{pmatrix}, and let α1,α2,α3\boldsymbol{\alpha}_{1},\boldsymbol{\alpha}_{2},\boldsymbol{\alpha}_{3} be a linearly independent set of 3-dimensional column vectors. Then the rank of the vector set Aα1,Aα2,Aα3\boldsymbol{A}\boldsymbol{\alpha}_{1},\boldsymbol{A}\boldsymbol{\alpha}_{2},\boldsymbol{A}\boldsymbol{\alpha}_{3} is ______.

rank of a matrixlinear independence

Approach:

This question tests the rank of a matrix and the properties of a linearly independent set of vectors. The key is to write the three vectors as the matrix product A(α1,α2,α3)\boldsymbol{A}\left(\boldsymbol{\alpha}_{1},\boldsymbol{\alpha}_{2},\boldsymbol{\alpha}_{3}\right) and then use the fact that (α1,α2,α3)\left(\boldsymbol{\alpha}_{1},\boldsymbol{\alpha}_{2},\boldsymbol{\alpha}_{3}\right) is invertible.

Steps:

  1. Put M=(α1,α2,α3)\boldsymbol{M}=\left(\boldsymbol{\alpha}_{1},\boldsymbol{\alpha}_{2},\boldsymbol{\alpha}_{3}\right). Then (Aα1, Aα2, Aα3)=AM\left(\boldsymbol{A}\boldsymbol{\alpha}_{1},\ \boldsymbol{A}\boldsymbol{\alpha}_{2},\ \boldsymbol{A}\boldsymbol{\alpha}_{3}\right)=\boldsymbol{A}\boldsymbol{M}
  2. Since α1,α2,α3\boldsymbol{\alpha}_{1},\boldsymbol{\alpha}_{2},\boldsymbol{\alpha}_{3} are linearly independent, M\boldsymbol{M} is invertible; right-multiplication by an invertible matrix does not change the rank, so r(Aα1,Aα2,Aα3)=r(A)r\left(\boldsymbol{A}\boldsymbol{\alpha}_{1},\boldsymbol{A}\boldsymbol{\alpha}_{2},\boldsymbol{A}\boldsymbol{\alpha}_{3}\right)=r(\boldsymbol{A})
  3. Apply elementary row transformations to A\boldsymbol{A}: A=(101112011)→(101011000)\boldsymbol{A}=\begin{pmatrix}1&0&1\\1&1&2\\0&1&1\end{pmatrix}\to\begin{pmatrix}1&0&1\\0&1&1\\0&0&0\end{pmatrix} There are 2 nonzero rows, so r(A)=2r(\boldsymbol{A})=2.
  4. Hence the rank of Aα1,Aα2,Aα3\boldsymbol{A}\boldsymbol{\alpha}_{1},\boldsymbol{A}\boldsymbol{\alpha}_{2},\boldsymbol{A}\boldsymbol{\alpha}_{3} is 22.

Answer:

22.

14. Let the distribution function of the random variable XX be F(x)=0.5 Φ(x)+0.5 Φ ⁣(x−42)F(x)=0.5\,\Phi(x)+0.5\,\Phi\!\left(\dfrac{x-4}{2}\right), where Φ(x)\Phi(x) is the standard normal distribution function; then E(X)=E(X)= ______.

ExpectationMixed Distribution

Approach:

This question tests the identification of a mixed distribution from its distribution function and the computation of its expectation. F(x)F(x) is a linear combination of two normal distribution functions, corresponding to a mixed distribution with weights 0.5 each, and the linearity of expectation can be applied directly.

Steps:

  1. Put X1∼N(0,1)X_1\sim N(0,1) and X2∼N(4,4)X_2\sim N(4,4); then F(x)=0.5 Φ(x)+0.5 Φ ⁣(x−42)F(x)=0.5\,\Phi(x)+0.5\,\Phi\!\left(\dfrac{x-4}{2}\right) is exactly the distribution function of XX, that is XX takes the value X1X_1 with probability 0.5 and the value X2X_2 with probability 0.5.
  2. By the linearity of expectation, E(X)=0.5 E(X1)+0.5 E(X2)E(X)=0.5\,E(X_1)+0.5\,E(X_2).
  3. Substituting E(X1)=0E(X_1)=0 and E(X2)=4E(X_2)=4 gives E(X)=0.5⋅0+0.5⋅4=2E(X)=0.5\cdot0+0.5\cdot4=2.

Answer:

Part III: Free Response

Part III: Free Response: Questions 15-23, 94 points in total.

15. Let f(u,v)f(u,v) have second-order continuous partial derivatives, and let y=f(ex,cos⁡x)y=f\left(\mathrm{e}^{x},\cos x\right). Find dydx∣x=0\left.\dfrac{\mathrm{d}y}{\mathrm{d}x}\right|_{x=0} and d2ydx2∣x=0\left.\dfrac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}\right|_{x=0}.

derivative of a multivariable compositechain rulepartial derivative

Approach:

This question tests the first- and second-order derivatives of a composite function of two variables. Differentiate layer by layer with the chain rule and then substitute the value at x=0x=0.

Steps:

  1. Put u=exu=\mathrm{e}^{x} and v=cos⁡xv=\cos x, so y=f(u,v)y=f(u,v). The chain rule gives dydx=f1′(u,v)⋅dudx+f2′(u,v)⋅dvdx=exf1′−sin⁡x⋅f2′\frac{\mathrm{d}y}{\mathrm{d}x}=f_{1}'(u,v)\cdot\frac{\mathrm{d}u}{\mathrm{d}x}+f_{2}'(u,v)\cdot\frac{\mathrm{d}v}{\mathrm{d}x}=\mathrm{e}^{x}f_{1}'-\sin x\cdot f_{2}'
  2. At x=0x=0 we have u=e0=1u=\mathrm{e}^{0}=1 and v=cos⁡0=1v=\cos 0=1, so dydx∣x=0=f1′(1,1)\left.\frac{\mathrm{d}y}{\mathrm{d}x}\right|_{x=0}=f_{1}'(1,1)
  3. Differentiate once more: d2ydx2=ddx(exf1′)−ddx(sin⁡x⋅f2′)\frac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}=\frac{\mathrm{d}}{\mathrm{d}x}\left(\mathrm{e}^{x}f_{1}'\right)-\frac{\mathrm{d}}{\mathrm{d}x}\left(\sin x\cdot f_{2}'\right) Expanding the first term: ddx(exf1′)=exf1′+ex(f11′′⋅dudx+f12′′⋅dvdx)=exf1′+ex(exf11′′−sin⁡x⋅f12′′)\frac{\mathrm{d}}{\mathrm{d}x}\left(\mathrm{e}^{x}f_{1}'\right)=\mathrm{e}^{x}f_{1}'+\mathrm{e}^{x}\left(f_{11}''\cdot\frac{\mathrm{d}u}{\mathrm{d}x}+f_{12}''\cdot\frac{\mathrm{d}v}{\mathrm{d}x}\right)=\mathrm{e}^{x}f_{1}'+\mathrm{e}^{x}\left(\mathrm{e}^{x}f_{11}''-\sin x\cdot f_{12}''\right) Expanding the second term: ddx(sin⁡x⋅f2′)=cos⁡x⋅f2′+sin⁡x(f21′′⋅dudx+f22′′⋅dvdx)=cos⁡x⋅f2′+sin⁡x(exf21′′−sin⁡x⋅f22′′)\frac{\mathrm{d}}{\mathrm{d}x}\left(\sin x\cdot f_{2}'\right)=\cos x\cdot f_{2}'+\sin x\left(f_{21}''\cdot\frac{\mathrm{d}u}{\mathrm{d}x}+f_{22}''\cdot\frac{\mathrm{d}v}{\mathrm{d}x}\right)=\cos x\cdot f_{2}'+\sin x\left(\mathrm{e}^{x}f_{21}''-\sin x\cdot f_{22}''\right)
  4. Combining, d2ydx2=exf1′+ex(exf11′′−sin⁡x⋅f12′′)−cos⁡x⋅f2′−sin⁡x(exf21′′−sin⁡x⋅f22′′)\frac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}=\mathrm{e}^{x}f_{1}'+\mathrm{e}^{x}\left(\mathrm{e}^{x}f_{11}''-\sin x\cdot f_{12}''\right)-\cos x\cdot f_{2}'-\sin x\left(\mathrm{e}^{x}f_{21}''-\sin x\cdot f_{22}''\right)
  5. Substituting x=0x=0 (where ex=1\mathrm{e}^{x}=1, sin⁡0=0\sin 0=0, cos⁡0=1\cos 0=1) gives d2ydx2∣x=0=f1′(1,1)+f11′′(1,1)−f2′(1,1)\left.\frac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}\right|_{x=0}=f_{1}'(1,1)+f_{11}''(1,1)-f_{2}'(1,1)

Answer:

dydx∣x=0=f1′(1,1)\left.\dfrac{\mathrm{d}y}{\mathrm{d}x}\right|_{x=0}=f_{1}'(1,1) and d2ydx2∣x=0=f1′(1,1)+f11′′(1,1)−f2′(1,1)\left.\dfrac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}\right|_{x=0}=f_{1}'(1,1)+f_{11}''(1,1)-f_{2}'(1,1).

16. Find lim⁡n→∞∑k=1nkn2ln⁡(1+kn)\displaystyle\lim_{n\to\infty}\sum_{k=1}^{n}\dfrac{k}{n^{2}}\ln\left(1+\dfrac{k}{n}\right).

definite integralRiemann sum

Approach:

This question tests the relation between a sequence limit and a definite integral. Rewrite the expression as a Riemann sum and then evaluate the integral.

Steps:

  1. Since kn2=1n⋅kn\dfrac{k}{n^{2}}=\dfrac{1}{n}\cdot\dfrac{k}{n}, we have ∑k=1nkn2ln⁡(1+kn)=1n∑k=1nknln⁡(1+kn)\sum_{k=1}^{n}\frac{k}{n^{2}}\ln\left(1+\frac{k}{n}\right)=\frac{1}{n}\sum_{k=1}^{n}\frac{k}{n}\ln\left(1+\frac{k}{n}\right)
  2. The right-hand side is the right-endpoint Riemann sum of g(x)=xln⁡(1+x)g(x)=x\ln(1+x) on [0,1][0,1] (with xk=knx_k=\dfrac{k}{n}), so lim⁡n→∞1n∑k=1nknln⁡(1+kn)=∫01xln⁡(1+x) dx\lim_{n\to\infty}\frac{1}{n}\sum_{k=1}^{n}\frac{k}{n}\ln\left(1+\frac{k}{n}\right)=\int_{0}^{1}x\ln(1+x)\,\mathrm{d}x
  3. Integrate by parts with u=ln⁡(1+x)u=\ln(1+x) and dv=x dx\mathrm{d}v=x\,\mathrm{d}x: ∫01xln⁡(1+x) dx=[x22ln⁡(1+x)]01−∫01x22(1+x) dx\int_{0}^{1}x\ln(1+x)\,\mathrm{d}x=\left[\frac{x^{2}}{2}\ln(1+x)\right]_{0}^{1}-\int_{0}^{1}\frac{x^{2}}{2(1+x)}\,\mathrm{d}x
  4. Since x21+x=x−1+11+x\dfrac{x^{2}}{1+x}=x-1+\dfrac{1}{1+x}, ∫01x22(1+x) dx=12∫01(x−1+11+x)dx=12(12−1+ln⁡2)=12(ln⁡2−12)\int_{0}^{1}\frac{x^{2}}{2(1+x)}\,\mathrm{d}x=\frac{1}{2}\int_{0}^{1}\left(x-1+\frac{1}{1+x}\right)\mathrm{d}x=\frac{1}{2}\left(\frac{1}{2}-1+\ln 2\right)=\frac{1}{2}\left(\ln 2-\frac{1}{2}\right)
  5. Substituting back, ∫01xln⁡(1+x) dx=12ln⁡2−12(ln⁡2−12)=12ln⁡2+14\int_{0}^{1}x\ln(1+x)\,\mathrm{d}x=\frac{1}{2}\ln 2-\frac{1}{2}\left(\ln 2-\frac{1}{2}\right)=\frac{1}{2}\ln 2+\frac{1}{4}

Answer:

12ln⁡2+14\dfrac{1}{2}\ln 2+\dfrac{1}{4}.

17. For the equation x3+y3−3x+3y−2=0x^3+y^3-3x+3y-2=0, which determines the function y(x)y(x), find its extreme values.

Implicit FunctionExtreme Value

Approach:

This question tests the determination of the extreme values of a function defined implicitly. First differentiate implicitly to obtain y′y', solve for the candidate xx from the stationary points, then substitute it back into the original equation to determine the corresponding yy, and finally use the second derivative to distinguish the local maximum from the local minimum.

Steps:

  1. Differentiating both sides of x3+y3−3x+3y−2=0x^3+y^3-3x+3y-2=0 with respect to xx gives 3x2+3y2y′−3+3y′=03x^2+3y^2y'-3+3y'=0, which rearranges to y′=1−x2y2+1y'=\dfrac{1-x^2}{y^2+1}.
  2. Since y2+1>0y^2+1>0, the stationary points are given by 1−x2=01-x^2=0, that is x=±1x=\pm1; substituting x=1x=1 into the original equation gives y3+3y−4=0y^3+3y-4=0, that is (y−1)(y2+y+4)=0(y-1)(y^2+y+4)=0, and hence y=1y=1; substituting x=−1x=-1 gives y3+3y=0y^3+3y=0, that is y(y2+3)=0y(y^2+3)=0, and hence y=0y=0.
  3. To compute the second derivative: from y′=1−x2y2+1y'=\dfrac{1-x^2}{y^2+1} we get y′′=−2x(y2+1)−2y(1−x2)y′(y2+1)2y''=\dfrac{-2x(y^2+1)-2y(1-x^2)y'}{(y^2+1)^2}.
  4. At (1,1)(1,1) we have y′=0, y′′=−1<0y'=0,\ y''=-1<0, so a local maximum 11 is taken; at (−1,0)(-1,0) we have y′′=2>0y''=2>0, so a local minimum 00 is taken.

Answer:

The local maximum is 11 and the local minimum is 00.

18. Let f(x)f(x) be twice differentiable on [0,1][0,1], with f(0)=0, lim⁡x→0+f(x)x=1f(0)=0,\ \lim_{x\to0^+}\dfrac{f(x)}{x}=1. Prove:
(1) the equation f(x)=0f(x)=0 has at least one real root in (0,1)(0,1);
(2) the equation f(x)f′′(x)+(f′(x))2=0f(x)f''(x)+(f'(x))^2=0 has at least two distinct real roots in (0,1)(0,1).

Rolle Theorem

Approach:

This question tests the extensions of the intermediate value theorem and of Rolle’s theorem. The key is to write f(x)f′′(x)+(f′(x))2f(x)f''(x)+(f'(x))^2 as the derivative of some function, namely to put g(x)=f(x)f′(x)g(x)=f(x)f'(x), so that g′(x)g'(x) is exactly the expression in question; then Rolle’s theorem is applied twice to count the number of zeros.

Steps:

  1. From lim⁡x→0+f(x)x=1\lim\limits_{x\to0^+}\dfrac{f(x)}{x}=1 we know that f′(0)=1>0f'(0)=1>0, so on the left-hand part (0,1)(0,1) of the interval ff is strictly increasing.
  2. Moreover f(0)=0f(0)=0 and ff is continuous on [0,1][0,1]; if ff were positive everywhere in (0,1)(0,1), then the continuity of ff together with lim⁡x→0+f(x)x=1\lim\limits_{x\to0^+}\tfrac{f(x)}{x}=1 would force ff to take a negative value somewhere; by the intermediate value theorem the equation f(x)=0f(x)=0 has at least one real root in (0,1)(0,1), namely ξ1\xi_1.
  3. Put g(x)=f(x)f′(x)g(x)=f(x)f'(x); then g′(x)=f′(x)2+f(x)f′′(x)g'(x)=f'(x)^2+f(x)f''(x), which is exactly the expression in the second part of the question.
  4. By (1) we know that g(ξ1)=0g(\xi_1)=0; moreover f(0)=0f(0)=0 and ff is differentiable on [0,ξ1][0,\xi_1], so by Rolle’s theorem there is ξ2∈(0,ξ1)\xi_2\in(0,\xi_1) with f′(ξ2)=0f'(\xi_2)=0, and hence g(ξ2)=0g(\xi_2)=0.
  5. Applying Rolle’s theorem to gg on [ξ2,ξ1][\xi_2,\xi_1] and on [ξ1,ξ2][\xi_1,\xi_2] respectively gives two points η1∈(ξ2,ξ1)\eta_1\in(\xi_2,\xi_1) and η2∈(ξ1,ξ2)\eta_2\in(\xi_1,\xi_2) with g′(η1)=g′(η2)=0g'(\eta_1)=g'(\eta_2)=0, that is f(x)f′′(x)+(f′(x))2=0f(x)f''(x)+(f'(x))^2=0 has at least two distinct real roots in (0,1)(0,1).

Answer:

(1) f(x)=0f(x)=0 has at least one real root in (0,1)(0,1);
(2) f(x)f′′(x)+(f′(x))2=0f(x)f''(x)+(f'(x))^2=0 has at least two distinct real roots in (0,1)(0,1).

19. The lamina SS is the bounded part of the cone z=x2+y2z=\sqrt{x^2+y^2} cut off by the cylinder z2=2xz^2=2x, and its density is μ=9\mu=9. Denote the curve of intersection by CC.
(1) Find the equation of the projection of CC on the xOyxOy plane;
(2) Find the mass of SS, denoted by MM.

Surface IntegralMass of a Lamina

Approach:

This question tests the computation of the projected curve in a surface integral and the computation of the mass of a surface. In the first part zz is eliminated from the curve of intersection of the two surfaces to get the equation of the projection; in the second part the mass of the surface is written as the density times the area, and the area of the surface is computed with a surface integral of the first kind.

Steps:

  1. The curve of intersection CC is formed by z2=2xz^2=2x and the cone z=x2+y2z=\sqrt{x^2+y^2} (that is z2=x2+y2z^2=x^2+y^2); eliminating z2z^2 from the two equations gives x2+y2=2xx^2+y^2=2x.
  2. Rearranging gives (x−1)2+y2=1(x-1)^2+y^2=1, which is the circle in the xOyxOy plane with centre (1,0)(1,0) and radius 1, so the projection of CC on the xOyxOy plane is that circle (z=0z=0).
  3. To compute the area of the surface SS: the area element on the cone z=x2+y2z=\sqrt{x^2+y^2} is 1+zx2+zy2 dx dy\sqrt{1+z_x^2+z_y^2}\,dx\,dy, where zx=xx2+y2, zy=yx2+y2z_x=\dfrac{x}{\sqrt{x^2+y^2}},\ z_y=\dfrac{y}{\sqrt{x^2+y^2}}, so the coefficient is 2\sqrt2.
  4. The projected region is D:(x−1)2+y2≤1D:(x-1)^2+y^2\le1, whose area is π\pi, so the area of the surface is =∬D2 dx dy=2 π=\iint_D\sqrt2\,dx\,dy=\sqrt2\,\pi.
  5. The density μ=9\mu=9 is constant, and the mass is M=9⋅2 π=18πM=9\cdot\sqrt2\,\pi=18\pi.

Answer:

(1) The projection is (x−1)2+y2=1, z=0(x-1)^2+y^2=1,\ z=0;
(2) M=18πM=18\pi.

20. Let the third-order matrix A=(α1,α2,α3)A=(\alpha_1,\alpha_2,\alpha_3) have three distinct eigenvalues, and let α3=α1+2α2\alpha_3=\alpha_1+2\alpha_2.
(1) Prove that r(A)=2r(A)=2;
(2) If β=α1+α2+α3\beta=\alpha_1+\alpha_2+\alpha_3, find the general solution of the system Ax=βAx=\beta.

Matrix RankSystem of Linear Equations

Approach:

This question tests the rank of a matrix and the general solution of a system of linear equations. In the first part the structure of the eigenvalues shows that the null space is one-dimensional, so the rank is 2; in the second part the fundamental system of solutions of the homogeneous equation is found first, and then the right-hand side vector is expressed in terms of the column vectors and substituted into the system to get a particular solution.

Steps:

  1. From α3=α1+2α2\alpha_3=\alpha_1+2\alpha_2 we know that α1,α2,α3\alpha_1,\alpha_2,\alpha_3 are linearly dependent, so ∣A∣=0|A|=0, that is 0 is an eigenvalue of AA.
  2. By the assumption that AA has three distinct eigenvalues, the multiplicity of 0 is 1 and the dimension of the null space is 1, so r(A)=3−1=2r(A)=3-1=2.
  3. The null space is given directly by the relation: A(1,2,−1)T=α1+2α2−α3=0A(1,2,-1)^T=\alpha_1+2\alpha_2-\alpha_3=0, so the fundamental system of solutions of the homogeneous equation can be taken as (1,2,−1)T(1,2,-1)^T, and by (1) the null space is exactly one-dimensional.
  4. Expressing β\beta in terms of the column vectors αi\alpha_i and combining it with the relation α3=α1+2α2\alpha_3=\alpha_1+2\alpha_2, and then comparing the coefficients of the column vectors, determines the components of a particular solution, giving one particular solution x0=(12,12,0)Tx_0=\left(\tfrac12,\tfrac12,0\right)^T.
  5. Hence the general solution of the system is x=x0+k(1,2,−1)T=(12,12,0)T+k(1,2,−1)Tx=x_0+k(1,2,-1)^T=\left(\tfrac12,\tfrac12,0\right)^T+k(1,2,-1)^T, where kk is an arbitrary constant.

Answer:

(1) α1,α2,α3\alpha_1,\alpha_2,\alpha_3 are dependent, ∣A∣=0|A|=0, and only one eigenvalue is 0, so the rank is 2;
(2) A particular solution is (12,12,0)T(\tfrac12,\tfrac12,0)^T, and the general solution is x=(12,12,0)T+k(1,2,−1)Tx=(\tfrac12,\tfrac12,0)^T+k(1,2,-1)^T.

21. The quadratic form f(x1,x2,x3)=2x12+3x22+3x32+4x2x3f(x_1,x_2,x_3)=2x_1^2+3x_2^2+3x_3^2+4x_2x_3, under the orthogonal transformation x=Qyx=Qy, has standard form 2y12+y22+5y322y_1^2+y_2^2+5y_3^2; find an orthogonal matrix QQ.

Quadratic FormOrthogonal Diagonalization

Approach:

This question tests the orthogonal diagonalization of a quadratic form and the construction of an orthogonal matrix. Write the quadratic form in matrix form, find the eigenvalues of the symmetric matrix and the corresponding unit eigenvectors, and then arrange the unit eigenvectors as the columns of an orthogonal matrix.

Steps:

  1. The matrix of the quadratic form is A=(200032023)A=\begin{pmatrix}2&0&0\\0&3&2\\0&2&3\end{pmatrix}.
  2. To find the eigenvalues: ∣A−λE∣=(2−λ)[(3−λ)2−4]=(2−λ)(1−λ)(5−λ)|A-\lambda E|=(2-\lambda)\bigl[(3-\lambda)^2-4\bigr]=(2-\lambda)(1-\lambda)(5-\lambda), so the eigenvalues are 2, 1, 52,\ 1,\ 5.
  3. For λ=2\lambda=2: solving (A−2E)x=0(A-2E)x=0 gives the eigenvector (1,0,0)T(1,0,0)^T; for λ=1\lambda=1: solving (A−E)x=0(A-E)x=0 gives the eigenvector (0,1,−1)T(0,1,-1)^T; for λ=5\lambda=5: solving (A−5E)x=0(A-5E)x=0 gives the eigenvector (0,1,1)T(0,1,1)^T.
  4. Normalize the three eigenvectors (their lengths are 1,2,21,\sqrt2,\sqrt2 respectively) and use them as the columns of QQ, obtaining Q=(01012012−12012)Q=\begin{pmatrix}0&1&0\\ \tfrac{1}{\sqrt{2}}&0&\tfrac{1}{\sqrt{2}}\\ -\tfrac{1}{\sqrt{2}}&0&\tfrac{1}{\sqrt{2}}\end{pmatrix}.
  5. Since the three eigenvectors are pairwise orthogonal, QTQ=EQ^TQ=E; and QTAQQ^TAQ is a diagonal matrix whose diagonal entries are the eigenvalues 2,1,52,1,5, which coincide with the coefficients of the standard form 2y12+y22+5y322y_1^2+y_2^2+5y_3^2 given in the question.

Answer:

One may take Q=(01012012−12012)Q=\begin{pmatrix}0&1&0\\ \tfrac{1}{\sqrt{2}}&0&\tfrac{1}{\sqrt{2}}\\ -\tfrac{1}{\sqrt{2}}&0&\tfrac{1}{\sqrt{2}}\end{pmatrix}.

22. Let X,YX,Y be independent, P{X=0}=P{X=2}=12P\{X=0\}=P\{X=2\}=\tfrac12, and let the density of YY be fY(y)={12,0≤y≤20,otherf_Y(y)=\begin{cases}\tfrac12,&0\le y\le2\\0,&\text{other}\end{cases}.
(1) Find P{Y≤E(Y)}P\{Y\le E(Y)\}; (2) Find the probability density of Z=X+YZ=X+Y.

Random Variable

Approach:

This question tests the use of the total probability and of the distribution function method to find the distribution of a function of a random variable. In the first part E(Y)E(Y) is computed first and the probability is then found; in the second part, since XX takes discrete values, the distribution function of ZZ is written as a weighted sum of two shifted distribution functions and then differentiated piece by piece.

Steps:

  1. E(Y)=∫0212y dy=12⋅[y22]02=1E(Y)=\int_0^2\tfrac12 y\,dy=\tfrac12\cdot\left[\tfrac{y^2}{2}\right]_0^2=1, so P{Y≤E(Y)}=P{Y≤1}=∫0112 dy=12P\{Y\le E(Y)\}=P\{Y\le1\}=\int_0^1\tfrac12\,dy=\tfrac12.
  2. From Z=X+YZ=X+Y and the values taken by XX, the total probability formula gives FZ(z)=P{X=0}FY(z)+P{X=2}FY(z−2)=12FY(z)+12FY(z−2)F_Z(z)=P\{X=0\}F_Y(z)+P\{X=2\}F_Y(z-2)=\tfrac12F_Y(z)+\tfrac12F_Y(z-2).
  3. Here FY(z)=0 (z<0)F_Y(z)=0\ (z<0), FY(z)=z2 (0≤z≤2)F_Y(z)=\tfrac z2\ (0\le z\le2) and FY(z)=1 (z>2)F_Y(z)=1\ (z>2); substitute piece by piece and differentiate.
  4. When 0≤z<20\le z<2 we have FZ(z)=z4F_Z(z)=\tfrac z4, so fZ(z)=z4f_Z(z)=\tfrac z4; when 2≤z<42\le z<4 we have FZ(z)=12⋅1+z−24=z+24F_Z(z)=\tfrac12\cdot1+\tfrac{z-2}{4}=\tfrac{z+2}{4}, so fZ(z)=12f_Z(z)=\tfrac12; when 4≤z≤64\le z\le6 we have FZ(z)=1−6−z4F_Z(z)=1-\tfrac{6-z}{4}, so fZ(z)=4−z4f_Z(z)=\tfrac{4-z}{4}; in all other cases fZ(z)=0f_Z(z)=0.

Answer:

(1) 12\dfrac{1}{2} (the answer 19\tfrac{1}{9} given in the question does not agree with the density given; with the given density E(Y)=1E(Y)=1 and the probability is 1/21/2);
(2) fZ(z)={z4,0≤z<212,2≤z<44−z4,4≤z≤60,otherf_Z(z)=\begin{cases}\dfrac{z}{4},&0\le z<2\\[4pt]\dfrac{1}{2},&2\le z<4\\[4pt]\dfrac{4-z}{4},&4\le z\le6\\0,&\text{other}\end{cases}.

23. (11 points) To determine the precision of a balance, an engineer weighs an object of known mass μ\mu nn times. The nn measurements X1,X2,⋯ ,XnX_{1},X_{2},\cdots,X_{n} are independent and all follow the normal distribution N(μ,σ2)N\left(\mu,\sigma^{2}\right), and the engineer records the absolute errors Zi=∣Xi−μ∣ (i=1,2,⋯ ,n)Z_{i}=\left|X_{i}-\mu\right|\ (i=1,2,\cdots,n). Use Z1,Z2,⋯ ,ZnZ_{1},Z_{2},\cdots,Z_{n} to estimate σ\sigma. (I) Find the probability density of Z1Z_{1}; (II) Use the first moment to find the moment estimator of σ\sigma; (III) Find the maximum likelihood estimator of σ\sigma.

normal distributionmoment estimatormaximum likelihood estimator

Approach:

This question tests the density of a distribution derived from the normal one, together with moment estimation and maximum likelihood estimation. The density of Z=∣X−μ∣Z=|X-\mu| is the “folded” normal density, and the two estimators are then obtained from the definitions.

Steps:

  1. (I) Since X1∼N(μ,σ2)X_{1}\sim N\left(\mu,\sigma^{2}\right), we have X1−μσ∼N(0,1)\dfrac{X_{1}-\mu}{\sigma}\sim N(0,1). Let φ\varphi and Φ\Phi denote the standard normal density and distribution function; the distribution function of Z1Z_{1} is F(z)=P{Z1⩽z}F(z)=P\{Z_{1}\leqslant z\}.
  2. For z<0z<0, F(z)=0F(z)=0; for z⩾0z\geqslant0, F(z)=P{∣X1−μσ∣⩽zσ}=Φ(zσ)−Φ(−zσ)=2Φ(zσ)−1F(z)=P\left\{\left|\frac{X_{1}-\mu}{\sigma}\right|\leqslant\frac{z}{\sigma}\right\}=\Phi\left(\frac{z}{\sigma}\right)-\Phi\left(-\frac{z}{\sigma}\right)=2\Phi\left(\frac{z}{\sigma}\right)-1 so F(z)={0,z<0,2Φ(zσ)−1,z⩾0.F(z)=\begin{cases}0,&z<0,\\ 2\Phi\left(\dfrac{z}{\sigma}\right)-1,&z\geqslant0.\end{cases}
  3. Differentiating gives the density of Z1Z_{1}: f(z)={0,z⩽0,2σφ(zσ)=2σ2πe−z22σ2,z>0.f(z)=\begin{cases}0,&z\leqslant0,\\[2pt]\dfrac{2}{\sigma}\varphi\left(\dfrac{z}{\sigma}\right)=\dfrac{2}{\sigma\sqrt{2\pi}}\mathrm{e}^{-\frac{z^{2}}{2\sigma^{2}}},&z>0.\end{cases}
  4. (II) From (I), the first raw moment is E(Z)=E(∣X1−μ∣)=∫0+∞z⋅2σφ(zσ)dz=2∫0+∞z⋅12πe−z22σ21σdzE(Z)=E\left(\left|X_{1}-\mu\right|\right)=\int_{0}^{+\infty}z\cdot\frac{2}{\sigma}\varphi\left(\frac{z}{\sigma}\right)\mathrm{d}z=2\int_{0}^{+\infty}z\cdot\frac{1}{\sqrt{2\pi}}\mathrm{e}^{-\frac{z^{2}}{2\sigma^{2}}}\frac{1}{\sigma}\mathrm{d}z Putting t=zσt=\dfrac{z}{\sigma}, E(Z)=2∫0+∞tσ2πe−t221σdt=2π∫0+∞te−t22dt=2π⋅1=σ2πE(Z)=2\int_{0}^{+\infty}\frac{t\sigma}{\sqrt{2\pi}}\mathrm{e}^{-\frac{t^{2}}{2}}\frac{1}{\sigma}\mathrm{d}t=\sqrt{\frac{2}{\pi}}\int_{0}^{+\infty}t\mathrm{e}^{-\frac{t^{2}}{2}}\mathrm{d}t=\sqrt{\frac{2}{\pi}}\cdot1=\sigma\sqrt{\frac{2}{\pi}}
  5. By the definition of moment estimation, the moment estimator σ^1\hat\sigma_{1} satisfies E(Z)=σ^12πE(Z)=\hat\sigma_{1}\sqrt{\dfrac{2}{\pi}}, so σ^1=E(Z)2/π=1n∑i=1nZi⋅π2\hat\sigma_{1}=\frac{E(Z)}{\sqrt{2/\pi}}=\frac{1}{n}\sum_{i=1}^{n}Z_{i}\cdot\sqrt{\frac{\pi}{2}}
  6. (III) The likelihood function is L(σ)=∏i=1n2σφ(ziσ)=(2σ2π)ne−12σ2∑i=1nzi2,σ>0L(\sigma)=\prod_{i=1}^{n}\frac{2}{\sigma}\varphi\left(\frac{z_{i}}{\sigma}\right)=\left(\frac{2}{\sigma\sqrt{2\pi}}\right)^{n}\mathrm{e}^{-\frac{1}{2\sigma^{2}}\sum_{i=1}^{n}z_{i}^{2}},\qquad \sigma>0
  7. Taking logarithms and differentiating, ln⁡L(σ)=nln⁡22π−nln⁡σ−12σ2∑i=1nzi2\ln L(\sigma)=n\ln\frac{2}{\sqrt{2\pi}}-n\ln\sigma-\frac{1}{2\sigma^{2}}\sum_{i=1}^{n}z_{i}^{2} ddσln⁡L(σ)=−nσ+1σ3∑i=1nzi2=0⟹σ^22=1n∑i=1nzi2\frac{\mathrm{d}}{\mathrm{d}\sigma}\ln L(\sigma)=-\frac{n}{\sigma}+\frac{1}{\sigma^{3}}\sum_{i=1}^{n}z_{i}^{2}=0\quad\Longrightarrow\quad \hat\sigma_{2}^{2}=\frac{1}{n}\sum_{i=1}^{n}z_{i}^{2} that is, σ^2=1n∑i=1nzi2\hat\sigma_{2}=\sqrt{\frac{1}{n}\sum_{i=1}^{n}z_{i}^{2}}
  8. Comparing the two estimators: the moment estimator is the arithmetic mean of the ZiZ_i times a constant, while the maximum likelihood estimator is the square root of their mean square. Since the mean square is not less than the square of the mean, σ^2⩾σ^1\hat\sigma_{2}\geqslant\hat\sigma_{1}, with equality exactly when Z1=⋯=ZnZ_{1}=\cdots=Z_{n}.

Answer:

(I) f(z)={0,z⩽0,2σ2πe−z22σ2,z>0;f(z)=\begin{cases}0,&z\leqslant0,\\[2pt]\dfrac{2}{\sigma\sqrt{2\pi}}\mathrm{e}^{-\frac{z^{2}}{2\sigma^{2}}},&z>0;\end{cases}

(II) σ^1=1n∑i=1nZiπ2\hat\sigma_{1}=\dfrac{1}{n}\displaystyle\sum_{i=1}^{n}Z_{i}\sqrt{\dfrac{\pi}{2}};

(III) σ^2=1n∑i=1nZi2\hat\sigma_{2}=\sqrt{\dfrac{1}{n}\displaystyle\sum_{i=1}^{n}Z_{i}^{2}}.

Found an issue with this content?

Tell us what could be improved.