Student Notes To Accompany MS4214: STATISTICAL INFERENCE
Student Notes To Accompany MS4214: STATISTICAL INFERENCE
Student Notes To Accompany MS4214: STATISTICAL INFERENCE
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Example 2.3 (Binomial sampling). The number of successes in n Bernoulli trials is a<br />
random variable R taking on values r = 0, 1, . . . , n with probability mass function<br />
� �<br />
n<br />
P (R = r) = θ<br />
r<br />
r (1 − θ) n−r .<br />
This is the exact same sampling scheme as in the previous example except that instead<br />
of observing the sequence y we only observe the total number of successes r. Hence the<br />
likelihood function has the form<br />
LR (θ|r) =<br />
� �<br />
n<br />
θ<br />
r<br />
r (1 − θ) n−r .<br />
The relevant mathematical calculations are as follows:<br />
� �<br />
n<br />
ℓR (θ|r) = ln + r ln(θ) + (n − r) ln(1 − θ)<br />
r<br />
S (θ) = r n − r<br />
+<br />
n 1 − θ<br />
⇒ ˆ θ = r<br />
n<br />
I (θ) = r n − r<br />
+<br />
θ2 (1 − θ) 2 > 0 ∀ θ<br />
E( ˆ θ) = E(r) nθ<br />
=<br />
n n = θ ⇒ ˆ θ unbiased<br />
Var( ˆ θ) = Var(r)<br />
n2 nθ(1 − θ)<br />
=<br />
n2 = θ(1 − θ)<br />
n<br />
E [I (θ)] = E(r) n − E(r) nθ n − nθ<br />
+ = +<br />
θ2 (1 − θ) 2 θ2 (1 − θ) 2<br />
=<br />
n<br />
θ(1 − θ) =<br />
�<br />
Var[ ˆ �−1 θ]<br />
and ˆ θ attains the Cramer-Rao lower bound (CRLB). �<br />
Example 2.4 (Germinating seeds). Suppose 25 seeds were planted and r = 5 seeds<br />
germinated. Then ˆ θ = r/n = 0.2 and Var( ˆ θ) = 0.2 × 0.8/25 = 0.0064. The relative<br />
likelihood is<br />
R1(θ) =<br />
� �5 � �20 θ 1 − θ<br />
.<br />
0.2 0.8<br />
Suppose 100 seeds were planted and r = 20 seeds germinated. Then ˆ θ = r/n = 0.2<br />
but Var( ˆ θ) = 0.2 × 0.8/100 = 0.0016. The relative likelihood is<br />
R2(θ) =<br />
� θ<br />
0.2<br />
� 20 � 1 − θ<br />
0.8<br />
� 80<br />
.<br />
Suppose 25 seeds were planted and it is known only that r ≤ 5 seeds germinated. In<br />
this case the exact number of germinating seeds is unknown. The information about θ<br />
is given by the likelihood function<br />
L (θ) = P (R ≤ 5) =<br />
21<br />
5�<br />
r=0<br />
� �<br />
25<br />
θ<br />
r<br />
r (1 − θ) 25−r .