Statistics

Chapter 05 – Joint Probability Distribution

Joint PMF Discrete

2
Joint PDF Continuous

3
 𝑋  time of connection
Joint PDF 𝑌  time of authorization

 Let the RV 𝑋 “denote the time until a computer server connects to

your machine” (in msec) and let 𝑌 “denote the time until the server
authorizes you as a valid user” (in msec). Each of these RVs measures
the wait from a common starting time and 𝑋 < 𝑌. Assume that the joint
PDF for 𝑋 and 𝑌 is:
→ 𝑓𝑋𝑌 𝑥, 𝑦 = 6 × 10−6 exp(−0.001𝑥 − 0.002𝑦) for 𝑥 < 𝑦
① Find 𝑃 0 ≤ 𝑥 ≤ ∞, 𝑥 ≤ 𝑦 ≤ ∞
② Find 𝑃 0 ≤ 𝑥 ≤ 1000, 𝑥 ≤ 𝑦 ≤ 2000
4
Joint PDF Cont’d. 1

∞ ∞ ∞ ∞
න න 𝑓𝑋𝑌 (𝑥, 𝑦)𝑑𝑦𝑑𝑥 = න න 6 × 10−6 𝑒 −0.001𝑥−0.002𝑦 𝑑𝑦 𝑑𝑥
−∞ −∞ 0 𝑥
∞ ∞
= 6 × 10−6 න න 𝑒 −0.002𝑦 𝑑𝑦 𝑒 −0.001𝑥 𝑑𝑥
0 𝑥
∞ ∞
𝑒 −0.002𝑥 −0.001𝑥
= 6 × 10−6 න 𝑒 𝑑𝑥 = 0.003 න 𝑒 −0.003𝑥 𝑑𝑥
0 0.002 0
1
= 0.003 =1
0.003

5
Joint PDF Cont’d. 2

1000 2000
𝑃(𝑋 ≤ 1000, 𝑌 ≤ 2000) = ∫0 ∫𝑥 𝑓𝑋𝑌 (𝑥, 𝑦)𝑑𝑦𝑑𝑥 =
−6 1000 2000 −0.002𝑦
6 × 10 ∫0 ∫𝑥 𝑒 𝑑𝑦 𝑒 −0.001𝑥 𝑑𝑥
𝑒 −0.002𝑥 − 𝑒 −4
−6 1000
= 6 × 10 ∫0 𝑒 −0.001𝑥 𝑑𝑥 =
0.002
1000 −0.003𝑥
0.003∫0 𝑒 − 𝑒 −4 𝑒 −0.001𝑥 𝑑𝑥
1 − 𝑒 −3 1 − 𝑒 21
= 0.003 − 𝑒 −4 =
0.003 0.001
0.003(316.738 − 11.578) = 0.915
6
Marginal Probability
Distributions (MPD)

→ The response time is the speed of page downloads. Let 𝑋 denote the
number of bars of service and let 𝑌 denote the response time for a
particular user and site.
7
MPD Cont’d. Discrete

 The MPD for 𝑋 is found by summing the probabilities in each column

whereas the MPD for 𝑌 is found by summing the probabilities in each

row.
𝑓𝑋 (3) = 𝑃(𝑋 = 3) = 𝑃(𝑋 = 3, 𝑌 = 1) + 𝑃(𝑋 = 3, 𝑌 = 2) +
𝑃(𝑋 = 3, 𝑌 = 3) + 𝑃(𝑋 = 3, 𝑌 = 4)
= 0.25 + 0.2 + 0.05 + 0.05 = 0.55

8
MPD Cont’d.

𝑓𝑌 (4)
𝑓𝑌 (3)
𝑓𝑌 (2)
𝑓𝑌 (1)

𝑓𝑋 (1) 𝑓𝑋 (2) 𝑓𝑋 (3)

9
MPD Cont’d. Continuous

10
MPD Cont’d.

 Let 𝑋 and 𝑌 be two jointly CRV with joint PDF:

3 2
𝑓𝑋𝑌 (𝑥, 𝑦) = ቐ𝑥 + 2 𝑦 0 ≤ 𝑥 ≤ 1, 0≤𝑦≤1
0 otherwise

○ Find 𝑓𝑋 (𝑥) for 𝑃 0 ≤ 𝑥 ≤ 1

○ Find 𝑓𝑌 (𝑦) for 𝑃 0 ≤ 𝑦 ≤ 1

11
MPD Cont’d.
∞
𝑓𝑋 (𝑥) = න 𝑓𝑋𝑌 (𝑥, 𝑦)𝑑𝑦 1
−∞
1
𝑓𝑋 (𝑥) = ቐ𝑥 + 2 0≤𝑥≤1
3 2 0 otherwise
=න 𝑥 + 𝑦 𝑑𝑦
0 2
1
1 3
= 𝑥𝑦 + 𝑦
2 0
1
=𝑥+
2

12
MPD Cont’d.
∞
𝑓𝑌 (𝑦) = න 𝑓𝑋𝑌 (𝑥, 𝑦)𝑑𝑥 3 2 1
−∞ 𝑓𝑌 (𝑦) = ቐ2 𝑦 + 2 0≤𝑦≤1
1
3 2 0 otherwise
=න 𝑥 + 𝑦 𝑑𝑥
0 2
1
1 2 3 2
= 𝑥 + 𝑦 𝑥
2 2 0
3 2 1
= 𝑦 +
2 2

13
Conditional Probability Distributions (CPD)

𝑌 𝑋 𝑋 𝑌
𝑓𝑌∣𝑋 (𝑌) 𝑓𝑋∣𝑌 (𝑋)

14
CPD Cont’d.

 Let 𝑋 and 𝑌 be two jointly CRV with joint PDF:

𝑥 2 𝑦 2 𝑥𝑦
𝑓𝑋𝑌 (𝑥, 𝑦) = ቐ 4 + 4 + 6 0 ≤ 𝑥 ≤ 1, 0≤𝑦≤2
0 otherwise
○ Find, the CPD of 𝑋 given 𝑌 = 𝑦;
1
‒ 𝑃 𝑋< ∣𝑌=𝑦
2

15
CPD Cont’d.
 Let’s find the marginal PDF of Y. we have:
1 2
𝑥 𝑦 2 𝑥𝑦
𝑓𝑌 (𝑦) = න + + 𝑑𝑥
0 4 4 6
3𝑦 2 + 𝑦 + 1
= , for 0 ≤ 𝑦 ≤ 2
12
 Thus, for 0 ≤ 𝑦 ≤ 2, we obtain:
𝑓𝑋𝑌 𝑥, 𝑦 3𝑥 2 + 3𝑦 2 + 2𝑥𝑦
𝑓𝑋∣𝑌 (𝑥 ∣ 𝑦) = = 2
, for 0 ≤ 𝑥 ≤ 1
𝑓𝑌 𝑦 3𝑦 + 𝑦 + 1

16
 CPD Cont’d.
 Thus, for 0 ≤ 𝑦 ≤ 2, we obtain:
3𝑥 2 + 3𝑦 2 + 2𝑥𝑦
𝑓𝑋∣𝑌 (𝑥 ∣ 𝑦) = ൞ 3𝑦 2 + 𝑦 + 1 0≤𝑥≤1
0 otherwise
1
1 3𝑥 2 + 3𝑦 2 + 2𝑥𝑦
2
𝑃 𝑋< ∣𝑌=𝑦 =න 2
𝑑𝑥
2 0 3𝑦 + 𝑦 + 1
1
1 2
= 2 𝑥 3 + 𝑦𝑥 2 + 3𝑦 2 𝑥 0
3𝑦 + 𝑦 + 1
3 2 𝑦 1
𝑦 + +
=2 2 4 8
3𝑦 + 𝑦 + 1 17
CPD Cont’d.
→ 𝑋 and 𝑌 denote the number of bars of signal strength and response
time, respectively. Then:
𝑃 𝑋 = 3, 𝑌 = 1
𝑃 𝑌=1 𝑋=3 = =
𝑃 𝑋=3
𝑓𝑋𝑌 (3,1)/𝑓𝑋 (3) = 0.25/0.55 = 0.454

𝑃 𝑋 = 3, 𝑌 = 2
𝑃 𝑌=2 𝑋=3 = =
𝑃 𝑋=3
𝑓𝑋𝑌 (3,2)/𝑓𝑋 (3) = 0.2/0.55 = 0.364
18
CPD Cont’d.

Conditional Probability Distributions

19
CPD Cont’d.

20
CPD Cont’d.

 The conditional mean is interpreted as the expected response time

given that one bar of signal is present. The conditional variance of 𝑌
given 𝑋 = 1 is 𝑓1∣1 (𝑌 ∣ 𝑋) 𝑓2∣1 (𝑌 ∣ 𝑋) 𝑓3∣1 (𝑌 ∣ 𝑋) 𝑓4∣1 (𝑌 ∣ 𝑋)

𝐸(𝑌 ∣ 1) = 𝜇𝑌∣1 = 1(0.05) + 2(0.1) + 3(0.1) + 4(0.75) = 3.55

𝑉 𝑌 1 = (1 − 3.55)2 0.05 + (2 − 3.55)2 0.1 +
(3 − 3.55)2 0.1 + (4 − 3.55)2 0.75 = 0.748
21
Independent Random Variables

22
CPD Cont’d.

 Suppose 𝑓𝑋𝑌 (𝑥, 𝑦) = 2 × 10−6 exp(−0.001𝑥 − 0.002𝑦) for 𝑥 ≥ 0

and 𝑦 ≥ 0. Show that 𝑋 and 𝑌 are independent.
∞
𝑓𝑋 (𝑥) = න 2 × 10−6 𝑒 −0.001𝑥−0.002𝑦 𝑑𝑦 = 0.001𝑒 −0.001𝑥 for 𝑥 > 0
0
∞
𝑓𝑌 (𝑦) = න 2 × 10−6 𝑒 −0.001𝑥−0.002𝑦 𝑑𝑥 = 0.002𝑒 −0.002𝑦 for 𝑦 > 0
0

 Therefore, 𝑓𝑋𝑌 𝑥, 𝑦 = 𝑓𝑋 𝑥 𝑓𝑌 (𝑦) for all 𝑥 and 𝑦, and 𝑋 and 𝑌 are

independent.

23
Covariance and Correlation

24
Covariance and Correlation

 The result is obtained by multiplying {𝑋 − 𝜇𝑋 } times {𝑌 − 𝜇𝑌 }, times

𝑓𝑋𝑌 (𝑥, 𝑦) for each point in the range (𝑋, 𝑌).

25
 Covariance and Correlation
→ 𝑋 and 𝑌 denote the number of bars of signal strength and response time,
respectively. Then: 𝑓𝑋 (1) 𝑓𝑋 (2) 𝑓𝑋 (3)
𝐸(𝑋) = 1(0.2) + 2(0.25) + 3(0.55) = 3.25
𝐸(𝑌) = 1(0.28) + 2(0.25) + 3(0.17) + 4(0.3) = 2.49

𝐸 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌 = 1 − 3.25 1 − 2.49 0.01 + 2 − 3.25 1 − 2.49 0.02 +

3 − 3.25 1 − 2.49 0.25
+ 1 − 3.25 2 − 2.49 0.02 + 2 − 3.25 2 − 2.49 0.03 +
3 − 3.25 2 − 2.49 0.2
+ 1 − 3.25 3 − 2.49 0.02 + 2 − 3.25 3 − 2.49 0.1 +
3 − 3.25 3 − 2.49 0.05
+ 1 − 3.25 4 − 2.49 0.15 + 2 − 3.25 4 − 2.49 0.1 +
3 − 2.35 4 − 2.49 0.05 = −0.5815
26
Covariance and Correlation

27
Covariance and Correlation

𝐸 𝑋𝑌 = 0 × 0 × 0.2 + 1 × 1 × 0.1 +
1 × 2 × 0.1 + 2 × 1 × 0.1 +
2 × 2 × 0.1 + 3 × 3 × 0.4 = 4.5

Joint Distribution for DRV 𝑋 and 𝑌

28
 Covariance and Correlation

𝐸(𝑋) = 0 × 0.2 + 1 × 0.2 + 2 × 0.2 + 3 × 0.4 = 1.8

𝑉(𝑋) = 0(0 − 1.8)2 × 0.2 + (1 − 1.8)2 × 0.2 +
(2 − 1.8)2 × 0.2 + (3 − 1.8)2 × 0.4 = 1.36
→ Because the MPD of 𝑌 is the same as for 𝑋, 𝐸(𝑌) = 1.8 and 𝑉 𝑌 =
1.36. Consequently,
𝜎𝑋𝑌 = 𝐸(𝑋𝑌) − 𝐸(𝑋)𝐸(𝑌) = 4.5 − (1.8)(1.8) = 1.26
→ Furthermore:
𝜎𝑋𝑌 1.26
𝜌𝑋𝑌 = = = 0.926
𝜎𝑋 𝜎𝑌 1.36 1.36

29