Data 2, 4, 6
μ̂ = 4
Sample mean.
For a normal sample, the maximum likelihood estimates are and (divisor , not ).
This page is easier if you are comfortable with:
The log-likelihood is . Maximising over gives the sample mean, and over gives the average squared deviation. For the data : , the squared deviations are , so , while the unbiased sample variance is (the sample variance).
At the maximum the log-likelihood simplifies to ; for the data above, . The MLE of the variance is biased by the factor but has lower mean squared error (mean squared error). The log-likelihood of the normal density uses the normal density.
This result justifies the use of the sample mean and the sums of squares in regression (the coefficient of determination) and ANOVA (ANOVA). The general method is maximum likelihood estimation, and the intervals for are in the confidence interval for a mean.
The best-fitting normal has the sample mean and the average squared deviation as its parameters.
The maximum is at the average squared deviation 5, not at 20/3.
These examples are worked by hand, step by step, so you can compare your own working.
Find for the data .
Find and the unbiased variance for the same data.
Find at the MLE for the same data.
A few situations come up again and again. Knowing them saves time and helps you spot answers that cannot be right.
μ̂ = 4
Sample mean.
(4 + 0 + 4)/3
2.6667.
8/2
4.
σ̂²/s² = (n − 1)/n
0.6667.
For a normal sample the MLEs are μ̂ = x̄ and σ̂² = (1/n)Σ(xᵢ − x̄)², using divisor n; the unbiased estimate uses n − 1.
Data 2, 4, 6. Find μ̂.
Find σ̂².
Find s².
Find σ̂²/s².
s² ✗
divide by n ✓
MLE is biased.
true mean unknown ✗
plug in x̄ ✓
Estimate.
Σ(x − x̄) ✗
Σ(x − x̄)² ✓
Sum of squares.
maximising separately fails ✗
μ̂ first then σ̂² ✓
Profile.
Try each one on paper, then open the answer. Use the solver above to see the full working.
4
5
6.6667
12
2.6667
0.8
Unlimited fresh questions with instant marking. Choose a level, work it out on paper, then type your answer.
Solve: …
x̄.
Σ(x − x̄)²/n.
It uses x̄ instead of μ.
Divide by n − 1.
maximum likelihood estimation.
More in Advanced statistics and stochastic processes and the formula sheets.