QUESTION 1

The random variable X is defined by the Probability Density Function (PDF):

f (x) = 7x6 , for 0 < x < 1

Calculation of Mean and Variance

1. Mean (µ = E[X]):
Z 1
7
E[X] = x · f (x)dx = = 0.875
0 8
2. Variance (σ 2 = Var[X]):
7
σ 2 = E[X 2 ] − (E[X])2 =
576

Convergence Analysis of the Five Sequences

The analysis of the sequences is grounded in the Weak Law of Large Numbers (WLLN) and the Central
Limit Theorem (CLT).

Sequence 1: Sn
• Sn does not converge to a constant value. It grows linearly with n.
• Explanation: The expected value is E[Sn ] = n · E[X] = 0.875n. Since E[X] > 0, the sum of a growing
number of positive random variables is expected to diverge to infinity as n → ∞.

Sequence 2: Sn /n
• Sn /n converges to the constant value 0.875.
• Explanation: This sequence is the sample mean (X̄n ). The plot demonstrates the Weak Law of Large
Numbers (WLLN), which states that the sample mean converges in probability to the population mean
µ, provided µ and σ 2 are finite:
Sn P
−
→ E[X] = 0.875
n

Sequence 3: Sn − n/2
• Sn − n/2 does not converge to a constant. It grows linearly with n.
• Explanation: The expected value is E[Sn − n/2] = n(E[X] − 1/2) = 0.375n. Since the expected value
diverges to infinity, the sequence itself diverges.

Sequence 4: (Sn − n/2)/n

• (Sn − n/2)/n converges to the constant value 0.375.
P
• Explanation: This sequence is Snn − 21 . By WLLN, Sn /n −
→ E[X] = 0.875. Therefore, the entire
sequence converges in probability to:
1
E[X] − = 0.875 − 0.5 = 0.375
2
This is the WLLN applied to the shifted sample mean.
√
Sequence 5: (Sn − n/2)/ n
• This sequence does not converge to a constant or zero. It exhibits increasing fluctuations centered
around a growing value. √
• Explanation: This sequence is normalized by n. The standard form of the Central Limit Theorem
−nµ d
(CLT) is Sσn √ n
−
→ N (0, 1). The given sequence is:

Sn − n/2 Sn − nµ √
√ = √ + 0.375 n
n n
√
The Sn√−nµ
n
term converges in distribution, but the second term 0.375 n diverges to ∞. This causes the
entire sequence to diverge in probability.

1
Comment on Speed of Convergence
• Graph 3 (Sn − n/2): This sequence diverges at a rate of O(n).
• Graph 4 ((Sn − n/2)/n): This sequence converges to 0.375. The rate at which √ the sample mean
fluctuations decay
√ around the limit is determined by the standard error, which is σ/ n. The convergence
speed is O(1/ n). This √ is significantly faster than the divergence
√ rate of Graph 3. √
• Graph 5 ((Sn − n/2)/ n): This sequence uses the n scaling characteristic of the CLT. The n
factor is the critical normalizing term required√to achieve convergence in distribution to a non-degenerate
Gaussian
√ limit (if centered correctly). Since n grows slower than n, the √ divergence rate of Graph 5,
O( n), is slower than the divergence rate of Graph 3, O(n), indicating the n scaling has had a significant
normalizing effect.

2
QUESTION 2
The underlying distribution is an Exponential distribution with parameter λ = 1.5. The standardized
sample mean Zn is calculated using the formula:

X̄n − µ
Zn = √
σ/ n

The Central Limit Theorem (CLT) states that for i.i.d. random variables {Xi } with finite mean µ and finite
variance σ 2 , the distribution of the standardized sample mean Zn converges in distribution to the standard
normal distribution, N (0, 1), as n → ∞:
d
Zn −
→ N (0, 1)
Convergence Observations (Steps 1-5)
• Small n (n = 5, 10): The distribution of the standardized means is visibly skewed to the right,
mirroring the skewness of the original Exponential distribution.
• Medium n (n = 20, 50): As n increases, the skewness rapidly decreases. The sampled density begins to
form a more symmetric bell shape, and its mean and variance approach that of the N (0, 1) curve.
• Large n (n = 100, 500, 1000): For large sample sizes, the histogram of the standardized means aligns
remarkably well with the theoretical Standard Normal (N (0, 1)) density curve. The convergence is near-
complete, demonstrating the asymptotic property of the CLT.

Convergence Conclusion
• The densities of the standardized means display a clear and rapid convergence towards the N (0, 1)
distribution as n becomes larger.
• The convergence theorem in place is the Central Limit Theorem (CLT).

3
QUESTION 3
The PDF for the Exponential distribution is f (x|λ) = λe−λx for x > 0.

Part I: Likelihood and Graphical MLE

1. Likelihood Function, L(λ|x): Pn
L(λ|x) = λn e−λ i=1 xi

2. Log-Likelihood Function, ℓ(λ|x):

n
X
ℓ(λ|x) = n ln(λ) − λ xi
i=1

3. Graphical MLE Estimate (The Peak): The theoretical MLE for the Exponential parameter is
λ̂ = 1/X̄. Graphically, this is the λ-value corresponding to the peak of the Likelihood and Log-Likelihood
curves.
• Observation from Plots (Task I): As n increases, the Likelihood and Log-Likelihood curves become
sharper and more peaked around the true value of λ = 1.5. This indicates that larger samples provide
much more precise information about the true parameter value. The peak of the function graphically
provides the MLE estimate λ̂.

Part II & III: Asymptotic Properties of MLE

The simulation of 100 MLEs for various n allows us to study the sampling distribution of λ̂.

Asymptotic Properties of MLE Discussion (Task IV)

Based on the simulated results, the asymptotic properties of the MLE are holding good:
1. Consistency: The MLE is consistent if it converges to the true parameter value as n → ∞. Finding:
As n increases, the Mean of the MLEs approaches the True λ (1.5). The density plots visibly show the
distribution concentrating around the true λ line, confirming consistency.
2. Asymptotic Efficiency / Asymptotic Normality: For large n, the sampling distribution of the
MLE approaches a normal distribution with variance Var(λ̂) ≈ λ2 /n. √
Finding: The Standard Deviation of the MLEs closely matches the Theoretical
√ Standard Deviation (λ/ n)
for large n. The decreasing SD confirms the rate of convergence is ∝ 1/ n. Furthermore, the density plots for
larger n become increasingly symmetric and bell-shaped, confirming asymptotic normality.

4
Figure 1: Simulation of Derived Sequences from X ∼ f (x) = 7x6 (N=1000)

5
Figure 2: Central Limit Theorem Simulation: Density of Standardized Sample Means (Zn ) vs. Standard
Normal Density (N (0, 1))

6
Figure 3: Task I: Likelihood Function L(λ|x) for Different Sample Sizes (Single Sample)

7
Figure 4: Task I: Log-Likelihood Function ℓ(λ|x) for Different Sample Sizes (Single Sample)

8
Figure 5: Task II & III: Sampling Distribution of MLE λ̂ (Density) for Various Sample Sizes (λ = 1.5, M = 100
Repetitions)

9
Figure 6: Summary Statistics (λ = 1.5, M = 100 Repetitions)

10