Scribd
Upload
9 views15 pages

ch5 Covariance 0

Uploaded by

arooshashmi204
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views15 pages

ch5 Covariance 0

Uploaded by

arooshashmi204
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Joint  

Probability  
5 Distributions  and  
Random  Samples

Week  5,  2011                  Stat  4570/5570              


Material  from  Devore’s  book  (Ed  8),  and  Cengage
Covariance

2
Covariance
When  two  random  variables  X  and  Y  are  not  independent,  
it  is  frequently  of  interest  to  assess  how  strongly  they  are  
related  to  one  another.

The  covariance  between  two  rv’s  X  and  Y  is


Cov(X,  Y)  =  E[(X  – µX)(Y  – µY)]

X,  Y  discrete

X,  Y  continuous

3
Covariance
If  both  variables  tend  to  deviate  in  the  same  direction  (both  
go  above  their  means  or  below  their  means  at  the  same  
time),  then  the  covariance  will  be  positive.  If  the  opposite  is  
true,  the  covariance  will  be  negative.

If  X  and  Y  are  not  strongly  related,  the  covariance  will  be  


near  0.

4
Covariance  shortcut
The  following  shortcut  formula  for  Cov(X,  Y)  simplifies  the  
computations.

Proposition
Cov(X,  Y)  =  E(XY)  – µX – µY

According  to  this  formula,  no  intermediate  subtractions  are  


necessary;;  only  at  the  end  of  the  computation  is  µX – µY
subtracted  from  E(XY).  

This  is  analogous  to  the  “shortcut” for  the  variance  


computation  we  saw  earlier.
5
Covariance
The  covariance  depends  on  both  the  set  of  possible  pairs  
and  the  probabilities  of  those  pairs.  

Below  are  examples  of  3  types  of  “co-­varying”:

(a)  positive  covariance;; (b)  negative   covariance;; (c)  covariance  near  zero
Figure  5.4 6
Example  1
An  insurance  agency  services  customers  who  have  both  a  
homeowner’s  policy  and  an  automobile  policy.  For  each  
type  of  policy,  a  deductible  amount  must  be  specified.

For  an  automobile  policy,  the  choices  are  $100  and  $250,  
whereas  for  a  homeowner’s  policy,  the  choices  are  $0,  
$100,  and  $200.

Suppose  an  individual  – Bob  -­-­ is  selected  at  random  from  


the  agency’s  files.  Let  X  =  his  deductible  amount  on  the  
auto  policy  and  Y  =  his  deductible  amount  on  the  
homeowner’s  policy.
7
Example  1 cont’d

Suppose  the  joint  pmf is  given  by  the  insurance  company  in  
the  accompanying  joint  probability  table:

What  is  the  covariance  between  X  and  Y?    

8
Correlation

9
Correlation
Definition
The  correlation  coefficient  of  X  and  Y,  denoted  by  
Corr(X,  Y),  ρX,Y,  or  just  ρ,  is  defined  by  

It  represents  a  “scaled” covariance  – correlation  ranges  


between  -­1  and  1.
10
Example
In  the  insurance  example,  what  is  the  correlation  between  
X  and  Y?

11
Correlation
Propositions

1. Cov(aX  +  b,  cY  +  d)  =  a  c  Cov  (X,  Y)

2. Corr(aX  +  b,  cY  +  d)  =  sgn(ac)  Corr(X,  Y)

3. For  any  two  rv’s  X  and  Y,        –1  ≤ Corr(X,  Y)  ≤ 1

4. ρ =  1  or  –1  iff  Y  =  aX  +  b  for  some  numbers  a  and  b  with  


a  ≠ 0.

12
Correlation
If  X  and  Y  are  independent,   then  ρ =  0,  but  ρ =  0  does  
not  imply  independence.

The  correlation  coefficient  ρ is  a  measure  of  the  linear  


relationship  between  X  and  Y,  and  only  when  the  two  
variables  are  perfectly  related  in  a  linear  manner  will  ρ be  
as  positive  or  negative  as  it  can  be.  

A  ρ less  than  1  in  absolute  value  indicates  only  that  the  


relationship  is  not  completely  linear,  but  there  may  still  be  a  
very  strong  nonlinear  relation.

13
Correlation
Also,  ρ =  0  does  not  imply  that  X  and  Y  are  independent,  
but  only  that  there  is  a  complete  absence  of  a  linear  
relationship.  

When  ρ =  0,  X  and  Y  are  said  to  be  uncorrelated.

Two  variables  could  be  uncorrelated  yet  highly  dependent


because  there  is  a  strong  nonlinear  relationship,  so  be
careful  not  to  conclude  too  much  from  knowing  that  ρ =  0.

14
Interpreting    Correlation

xkcd.com/552/

15

You might also like