Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

STATISTICS 2016-17 WORKING SESSION 2

1. (2 pts.) A data transmission channel has a binary error rate of 0.15 (i.e., the probability that a bit is
corrupted during transmission is 0.15). If one tries to transmit 10 bits:

(a) Calculate the probability that at least 2 bits are corrupted during transmission.
(b) Calculate the probability that the first corrupt bit is the fifth one.
(c) Calculate the most likely amount of corrupt bits.
(d) Which bit is the most probable to be the first corrupt bit?

Solution:
P(corrupt)=0.15, P(not corrupt)=0.85

(a) binomial:

P (X 2) = 1 [P (X = 0) + P (X = 1)] = 1 (0.85)10 + 10 (0.85)9 0.15


 

= 1 (0.197 + 0.347) = 1 0.544 = 0.456

(b) geometric: P (X1 = 5) = (0.85)4 0.15 = 0.078


(c) binomial: NPmax = 0.15 10 = 1.5 verify X = 1 and X = 2.
P (X = 1) = 0.347
  (see above)
10 10!
P (X = 2) = (0.15)2 (0.85)8 = (0.15)2 (0.85)8 = 0.276
2 2! 8!
P (X = 1) > P (X = 2), conclusion: NPmax = 1

(d) The first (as always in the geometric model)

1
2. (2 pts.) Leo Messi scores, on average in the current season, one goal every 80 minutes. Assuming that a
game lasts exactly 90 minutes, calculate the probability that:

(a) Messi scores at least 3 goals in the next 2 games.


(b) (at least) 3 games go by without Messi scoring any goal, knowing that he has scored 2 goals in the
previous game.

Solution: POISSON
= 1/80 mins1
180 9
(a) 2 games=180 mins = t = = = 2.25
80 4

P (X 3) = 1 [P (X = 0) + P (X = 1) + P (X = 2)]
" 0 1 2 #
9 9 9
= 1 e9/4 4
+ 4
+ 4
0! 1! 2!
 
9 81 185 9/4
= 1 e9/4 1 + + =1 e
4 32 32
= 1 0.105 5.78 = 0.391

[Recall that x0 = 1 (x 6= 0) and that 0! = 1]


270 27
(b) 3 games=270 mins = t = = = 3.375
80 8
knowing that he has scored 2 goals in the previous game: IRRELEVANT, Poisson has no memory.
No goals during 3 games:
27 0

27/8
P (X = 0) = e 8
= e27/8 = 0.0342
0!

2
3. (6 pts.) In the first year of Bio-informatic engineering, there are two Calculus professors: Paco and Marta.
In the academic year 2015-16, a first group of 14 students has Paco as teacher. A second group of 15 has
Marta. The final exam is common for both groups and the final grades (out of 10) are summarized in the
following table.

Group I (Paco) Group II (Marta)


Xx = 5.90 Xx = 6.78
x2i = 512.82 x2i = 706.47
i i

Both Paco and Marta have been working as Calculus teachers for many years, and have had hundreds of
students in total. [Note: You may disregard the fact that the grades should necessarily lie between 0 and
10. Also, you may assume that the grades form a continuous spectrum.]

(a) (1 pt.) Use the data above to give unbiased point estimators for the mean and standard deviation of
the grades that Paco and Marta each give on Calculus throughout the years.
(b) (2 pts.) Use a confidence interval test to verify whether these data support, with a significance of 5%
and of 10%, the hypothesis that Paco and Marta grade the Calculus exams differently.
(c) (2 pts.) Apply a p-test to these same data and verify the same hypothesis at a significance level of 1;
5 and 10%.
(d) (0.5 pts.) Assume that exactly the same data had been obtained from two groups of 40 students each.
Indicate what the
essential change will be in all the previous calculations (apart from the obvious
fact of adapting n wherever this appears).
(e) (0.5 pts.) At a significance level of 5%, which one of the following options best expresses the conclu-
sion:

3
i. It is 95% correct that there is no statistically relevant difference between how Paco and Marta
grade the Calculus exams.
ii. With 95% of confidence, we can assert that there is no statistically relevant difference between
how Paco and Marta grade the Calculus exams.
iii. We cannot assert with 95% confidence that there is a statistically significant different between
how Paco and Marta grade the Calculus exams. Therefore we accept the hypothesis that there
is no such difference, with 5% significance.
iv. It is 95% correct that there is a statistically significant difference between how Paco and Marta
grade the Calculus exams.
v. With 95% of confidence, we can assert that there is a statistically relevant difference between
how Paco and Marta grade the Calculus exams.

Solution:

(a) (1 pt.) Unbiased estimators:



=x

n n
!
1 X 1 X n

= sn1 where s2n1 = )2 =
(xi x x2i 2
x
n1 n1 n1
i=1 i=1
So, for Group I (Paco):
1 = 5.90
1 14
12 =
512.82 (5.90)2 = 1.96
1 = 1.4
13 13
For Group 2 (Marta):
2 = 6.78
1 15
22 = 706.47 (6.78)2 = 1.21
2 = 1.1
14 14
(b) (2 pts.) CI test
H0 : 1 2 = 0
H1 : 1 2 6= 0
Difference distribution :
x
= x 2 = 0.88
s1
1.96 1.21
r
12
22

= + = + = 0.22 = 0.47
n1 n2 14 15
n1 , n2 < 30 and unknown Student-t distribution, 2-sided

4
= 5%: /2 = 0.025, = n1 + n2 2 = 27. t/2, = t0.025,27 = 2.052
t/2,
= 2.052 0.47 = 0.964
Confidence Interval CI : ( x1 x
2 ) t/2,

0.88 0.96 1 2 0.88 + 0.96


1.84 1 2 0.08 at a CL of 95%

1 2 = 0 CI, so we ACCEPT H0 at = 5% .
= 10%, i.e. /2 = 0.05
The only thing that changes is t: t/2, = t0.05,27 = 1.703
t/2,
= 1.703 0.47 = 0.80
Confidence Interval CI : ( x1 x
2 ) t/2,

0.88 0.80 1 2 0.88 + 0.80


1.68 1 2 0.08 at a CL of 90%

1 2 = 0
/ CI, so we REJECT H0 / ACCEPT H1 at = 10% .
(c) (2 pts.) p-test:
1 x
x 2 0.88
t= = = 1.872; = 27.

0.47
For = 0.05: t = 1.703. For = 0.025: t = 2.052.
0.05 + 0.025
t = 1.872 lies right in the middle, so = = 0.0375
2
2-sided p = 2 = 0.075 = 7.5% .

p = max accept H0 , so:

ACCEPT H0 for = 1%, = 5%


REJECT H0 for = 10%
(d) (0.5 pts.) For n > 30, one uses a normal distribution (Z) instead of student-t.

(e) (0.5 pts.) c) .


There is not enough statistical evidence to reject H0 with 95% confidence, but that does not mean
that you really trust H0 (and certainly not with 95% confidence).

You might also like