Upload
betty-sutton
View
213
Download
0
Embed Size (px)
Citation preview
Estimation of covariance matrix under informative
sampling
Julia AruUniversity of Tartu and Statistics Estonia
Tartu, June 25-29, 2007
Tartu, June 26-29, 2007 2
Outline
• Informative sampling• Population and sample distribution• Multivariate normal distribution and exponential
inclusion probabilities• Conclusions for normal case• Simulation study
Tartu, June 26-29, 2007 3
Informative sampling
• Probability that an object belongs to the sample depends on the variable we are interested in
• For example: while studying income we see that people with higher income are not keen to respond
• Under informative sampling sample distribution of variable(s) of interest differs from that in population
Tartu, June 26-29, 2007 4
Population and sample distribution
• Vector of study variables • Population distribution • Sample distribution
),...,,( 21 kyyyy)(ypf
)(
)()|()1|()(
p
spps E
fEIff
yyyy
Tartu, June 26-29, 2007 5
MVN case (1)
• Population distribution: multivariate normal with parameters µ and Σ:
• Inclusion probabilities:
• Matrix A is symmetrical and such that
is positive-definite
)''exp()|( ybAyyy cEp
1
/ 2
1 1( | , ) exp ( ) ' ( )
2(2 ) | |p kf
y μ Σ y μ Σ y μ
Σ
1 1( 2 ) Σ A
Tartu, June 26-29, 2007 6
MVN case (2)
• Sample distribution is then again normal with parameters
1 1 1' ( ' ')( 2 ) λ μ Σ b Σ A
1 1( 2 ) Ω Σ A
Tartu, June 26-29, 2007 7
Conclusions for MVN case• If variables are independent in the
population (Σ is diagonal) then independence is preserved only in the case when matrix A is also diagonal
1 1 1' ( ' ')( 2 ) λ μ Σ b Σ A1 1( 2 ) Ω Σ A
• Matrix A can be chosen to make variables independent in the sample or dependence structure to be very different from that in the population
Tartu, June 26-29, 2007 8
Simulation study (1)
• Population is bivariate standard normal with correlation coefficient r :
• Inclusion probabilities:
• Repetitions: 1000, population size: 10000, sample size: 1000
)'0 ,0(μ
R̂
1
1
r
rΣ
12/1
2/11A )'1 ,1(b
Tartu, June 26-29, 2007 9
Simulation study (2)
R̂r R
-1 -1 -1
-0.8 -0.26 -0.25
-0.6 0.02 0.01
-0.4 0.16 0.17
-0.2 0.26 0.27
0 0.33 0.33
0.2 0.40 0.39
0.4 0.46 0.46
0.6 0.54 0.53
0.8 0.67 0.68
1 1 1
Tartu, June 26-29, 2007 10
Thank you!
Tartu, June 26-29, 2007 11
Exponential family (1)
• Population distribution belongs to expontial family
• With canonocal representation
• And inclusion probabilities have the form
1
( | ) ( ) exp ( ) ( ) ( )d
p i ii
f h g T B
y θ y θ y θ
* *
1
( | ) ( ) exp ( ) ( )d
p i ii
f h T B
y η y y η
1
( | ) exp ( )d
p i ii
E r pT
y y
Tartu, June 26-29, 2007 12
Exponential family (2)
• Then sample distribution belonds to the same family of distributions with canonical parameters
*i i ip