Estimators, standard error and bias

Estimators, standard error and bias Test

Spec 5.1 · Edexcel A-Level Further Maths: Subtopic test

Studying this course? Keep your progress in one place.

Two ways to do this test

Question by question

One question at a time. Each answer is marked the moment you hand it in, so you learn as you go. Drago helps if you get stuck. All 10 questions, or just 5.

Exam mode

The whole paper in one go, timed like the real exam. Nothing is marked until you hand it in, then you get every mark explained.

10 questions, 27 marks

Edexcel A-Level Further Maths

Estimators, standard error and bias

Total 27 marks

Name

Class

Date

  1. 1
    X1X_1, X2X_2, X3X_3 is a random sample from a population with mean μ\mu and variance σ2\sigma^2. The statistic U=X1+2X2+3X36U=\frac{X_1+2X_2+3X_3}{6} is proposed as an estimator of μ\mu.
    (a)
    Which statement about UU is correct?
    [1 mark]
    • AUU is unbiased, because E(U)=μ\mathrm{E}(U)=\mu
    • BUU is biased, because the weights are not all equal
    • CUU is biased, because E(U)=6μ\mathrm{E}(U)=6\mu
    • DUU is unbiased only if μ=0\mu=0
    Show the mark scheme

    Mark scheme

    Answer: A
    (b)
    Find Var(U)\mathrm{Var}(U).
    [1 mark]
    • Aσ23\frac{\sigma^2}{3}
    • Bσ26\frac{\sigma^2}{6}
    • C7σ23\frac{7\sigma^2}{3}
    • D7σ218\frac{7\sigma^2}{18}
    Show the mark scheme

    Mark scheme

    Answer: D
    (c)
    The sample mean Xˉ=X1+X2+X33\bar X=\frac{X_1+X_2+X_3}{3} is also an estimator of μ\mu. Compare UU and Xˉ\bar X and state, with a reason, which you would use.
    [2 marks]
    Show the mark scheme

    Mark scheme

    B1: both are unbiased, E(U)=E(Xˉ)=μ\mathrm{E}(U)=\mathrm{E}(\bar X)=\mu
    B1:
    Var(Xˉ)=σ23<7σ218=Var(U)\mathrm{Var}(\bar X)=\frac{\sigma^2}{3}<\frac{7\sigma^2}{18}=\mathrm{Var}(U), so use Xˉ\bar X because it has the smaller variance
    Accept:
    618σ2<718σ2\frac{6}{18}\sigma^2<\frac{7}{18}\sigma^2
    Do not accept:
    Xˉ\bar X preferred with no variance comparison

    Total for question 1: 4 marks

  2. 2
    A random sample of 1010 observations of a quantity xx from a population with unknown mean and variance gives ∑x=124.0\sum x=124.0 and ∑x2=1582.6\sum x^2=1582.6.
    (a)
    Calculate an unbiased estimate of the population variance.
    [1 mark]
    • A4.54.5
    • B55
    • C2.242.24
    • D4545
    Show the mark scheme

    Mark scheme

    Answer: B
    (b)
    Calculate the estimated standard error of the sample mean.
    [1 mark]
    • A0.2240.224
    • B2.242.24
    • C0.7070.707
    • D0.50.5
    Show the mark scheme

    Mark scheme

    Answer: C
    (c)
    A larger sample of 4040 observations has the same sample variance as in part (a). Calculate the estimated standard error of its mean and state what this shows about Xˉ\bar X as an estimator of the population mean.
    [2 marks]
    Show the mark scheme

    Mark scheme

    B1: 540=0.354\sqrt{\frac{5}{40}}=0.354
    B1: standard error is halved, so
    Xˉ\bar X varies less from sample to sample / is a more precise estimator
    Accept:
    24\frac{\sqrt2}{4}
    Accept: variance of the estimator decreases as
    nn increases

    Total for question 2: 4 marks

  3. 3
    X1,X2,…,XnX_1,X_2,\ldots,X_n is a random sample from a population with mean μ\mu and variance σ2\sigma^2. Two estimators of σ2\sigma^2 are S2=1n−1∑(Xi−Xˉ)2S^2=\frac{1}{n-1}\sum(X_i-\bar X)^2, which is known to be unbiased, and V=1n∑(Xi−Xˉ)2V=\frac{1}{n}\sum(X_i-\bar X)^2.
    (a)
    Show that VV is a biased estimator of σ2\sigma^2, and state the bias.
    [3 marks]
    Show the mark scheme

    Mark scheme

    M1: V=n−1nS2V=\frac{n-1}{n}S^2
    M1:
    E(V)=n−1nE(S2)=n−1nσ2\mathrm{E}(V)=\frac{n-1}{n}\mathrm{E}(S^2)=\frac{n-1}{n}\sigma^2
    A1: bias
    =E(V)−σ2=−σ2n=\mathrm{E}(V)-\sigma^2=-\frac{\sigma^2}{n}, which is not zero, so VV is biased
    Accept:
    VV underestimates σ2\sigma^2 on average
    (b)
    A sample of size 88 has ∑(xi−xˉ)2=56\sum(x_i-\bar x)^2=56. Calculate the estimates of σ2\sigma^2 given by S2S^2 and by VV. Hence state, as a percentage of σ2\sigma^2, the size of the bias of VV for n=8n=8, and say which estimator you would use.
    [4 marks]
    Show the mark scheme

    Mark scheme

    B1: s2=567=8s^2=\frac{56}{7}=8
    B1:
    v=568=7v=\frac{56}{8}=7
    A1: bias
    =−σ28=-\frac{\sigma^2}{8}, i.e. VV underestimates by 12.5%12.5\% of σ2\sigma^2 on average
    R1: use
    S2S^2 because it is unbiased
    Accept:
    −12.5%-12.5\%

    Total for question 3: 7 marks

  4. 4
    X1,X2,X3,X4X_1,X_2,X_3,X_4 is a random sample from a population with mean μ>0\mu>0 and variance σ2\sigma^2. Three estimators of μ\mu are A=X1+X2+X3+X44A=\frac{X_1+X_2+X_3+X_4}{4}, B=X1+X22B=\frac{X_1+X_2}{2} and C=X1+X2+X3+X45C=\frac{X_1+X_2+X_3+X_4}{5}.
    (a)
    (i) Show that AA is an unbiased estimator of μ\mu.
    (ii) Show that
    BB is unbiased.
    (iii) Find
    E(C)\mathrm{E}(C) and hence the bias of CC, and state whether CC over- or underestimates μ\mu.
    [6 marks]
    Show the mark scheme

    Mark scheme

    M1: (i) E(A)=14(μ+μ+μ+μ)\mathrm{E}(A)=\frac14(\mu+\mu+\mu+\mu)
    A1: (i)
    =μ=\mu, so AA is unbiased
    AG

    B1: (ii)
    E(B)=12(μ+μ)=μ\mathrm{E}(B)=\frac12(\mu+\mu)=\mu, so BB is unbiased
    M1: (iii)
    E(C)=4μ5\mathrm{E}(C)=\frac{4\mu}{5}
    A1: (iii) bias
    =E(C)−μ=−μ5=\mathrm{E}(C)-\mu=-\frac{\mu}{5}
    R1: (iii)
    CC underestimates μ\mu (since μ>0\mu>0)
    (b)
    Find the variance of each of AA, BB and CC, and hence evaluate which estimator you would recommend for μ\mu, justifying your choice.
    [6 marks]
    Show the mark scheme

    Mark scheme

    M1: Var(A)=116(4σ2)\mathrm{Var}(A)=\frac{1}{16}(4\sigma^2) using independence and Var(aX)=a2Var(X)\mathrm{Var}(aX)=a^2\mathrm{Var}(X)
    A1:
    Var(A)=σ24\mathrm{Var}(A)=\frac{\sigma^2}{4}
    B1:
    Var(B)=σ22\mathrm{Var}(B)=\frac{\sigma^2}{2}
    B1:
    Var(C)=4σ225\mathrm{Var}(C)=\frac{4\sigma^2}{25}
    R1:
    AA and BB are both unbiased, and AA has the smaller variance, so AA is better than BB
    R1:
    CC has the smallest variance but is biased, so AA is recommended, as the unbiased estimator with the smaller variance
    Accept:
    CC rejected because its bias is systematic

    Total for question 4: 12 marks

End of questions

Subtopic tests