Your SlideShare is downloading. ×
  • Like
D03402029035
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Now you can save presentations on your phone or tablet

Available for both IPhone and Android

Text the download link to your phone

Standard text messaging rates apply
Published

The International Journal of Engineering & Science is aimed at providing a platform for researchers, engineers, scientists, or educators to publish their original research results, to exchange new …

The International Journal of Engineering & Science is aimed at providing a platform for researchers, engineers, scientists, or educators to publish their original research results, to exchange new ideas, to disseminate information in innovative designs, engineering experiences and technological skills. It is also the Journal's objective to promote engineering and technology education. All papers submitted to the Journal will be blind peer-reviewed. Only original articles will be published.
The papers for publication in The International Journal of Engineering& Science are selected through rigorous peer reviews to ensure originality, timeliness, relevance, and readability.

Published in Engineering , Technology , Business
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
47
On SlideShare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
1
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. The International Journal Of Engineering And Science (IJES) || Volume || 3 || Issue || 4 || Pages || 29-35 || 2014 || ISSN (e): 2319 – 1813 ISSN (p): 2319 – 1805 www.theijes.com The IJES Page 29 A note on estimation of parameters of multinormal distribution with constraints 1, Dr. Parag B. Shah, 2, C. D. Bhavsar Dept. of Statistics, H.L. College of Commerce, Ahmedabad-380 009. INDIA Dept. of Statistics, School of Science, Gujarat University, Ahmedabad-380 009. INDIA -------------------------------------------------------ABSTRACT--------------------------------------------------- Mardia et al (1989) considered the problem of estimating the parameters of nonsingular multivariate normal distribution with certain constraints. Nagmur (2003) considered the problem of estimating the mean sub-vector of non-sigular multivariate normal distribution with certain constraints. In this paper we try to estimate mean sub-vector under some different constraints and submatrix of ∑ with certain constraints for a nonsingular multivariate normal distribution. Keywords: Likelihood Function, Maximum Likelihood Estimator, Non-singular Multivariate Normal Distribution, Constraints. ---------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 31 March 2014 Date of Publication: 15 April 2014 ----------------------------------------------------------------------------------------------------------------- I. INTRODUCTION Mardia et al (1989) considered the estimation of parameters of non-singular multivariate normal distribution with and without constraints. Nagnur (2003) tried to obtain the Mle of sub mean vector of the distribution which can be useful in some practical problems. Mardia et al (1989) considered two type of constraints on the mean vector  . i. = k i.e.  is known to be proportional to a known vector  . For example, the elements of x could represent a sample of repeated measurements, in which case  = k1 ii. Another type of constraint is R = r , where R and r are pre-specified The first type of constraint was considered by Nagnur (2003) for sub mean vector of the distribution. In this paper, we consider type of constraint for sub mean vector and give the explicit expression for the estimators. Mardia et al (1989) also considered constraint on variance-covariance matrix , viz =k0 where 0 is known. We consider constraint on sub matrix of  and obtain its estimator with constraints. 2. ML Estimator of µ : Let 1x , 2x ,…, nx be n iid observation from Np  ,  population, where  is positive definite matrix. Suppose that x ,  and  are partitioned as follows : 1 2 x x x         , 1 2            , 11 12 21 22           11  : r x r, 22  : s x s with r + s = p. Let x and   1 n i i x x xixS   ' be the mean vector and SSP matrix based on the sample observations 1 2 , ,........... n x x x . The sample mean vector and SSP matrix is partitioned as
  • 2. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 30 1 2 x x x        11 12 21 22 S S S S S         where 11 S : r x r , 22 S : s x s ,with r + s = p. Our problem is to estimate  : r  1 under the constraint R = r , where R and r are pre-specified. Maximizing the log likelihood subject to this constraint may be achieved by augmenting the log likelihood with a Lagrangian expression, i.e. we maximize L+ = L - n ' (Rµ - r) (2.1) where  is a vector of Lagrangian multipliers and L is given by    1 11 22 22 log 2 logn nnp L tr S x x              ' (2.2) Case-1 ( Known ) With  assumed to be known, to find m.l.e.'s of 1  we are required to find  for which the solution to 1 0L      satisfies the constraint R 1  = r. Observe that L  can be expressed as    1 111 2 1 1 log 2 log 2 2 2 }{np n n L tr S x x '                       1 2 2 2 ' 12 22 1 2 2 2 1 ' 2 ' 2 n x x x x n R r                            (2.3) Now    11 12 21 21 1 ' 0L n x n x nR              l (2.4) and    22 21 2 1 2 2 1 0L n x n x             (2.5) From (2.5) we get      1 12 2 1 21 22 x x         (2.6) Substituting for  2 2 x  in (2.4), we get
  • 3. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 31     111 12 22 21 1111 ' 0n x n x nR                 Since   111 12 22 21 1 11 . .          ., we get  1 111 1 ' ..n x nR     (2.7) Thus  1 1 x  = 11 'R  , Pre-multiplying by R gives    111 'Rx r R R    if the constraint 1 R = r is to be satisfied. Thus, we take,     1 11 1 'R R Rx r     ,so 1 111 'x R      (2.8) From (6), the ML estimator of 2  is      122 21 2 1 12 x x         (2.9) Case 2 ( ∑ unknown) : When the covariance matrix ∑ is not known, we have to estimate 1 2 ,  and ∑ using the likelihood function (2.3) The ML estimator of ∑ is  * S n   , where * S =  n i 1 'i ix x                    (2.10) and    is the ML estimator of  under the restriction R 1  = r. To estimate ( 1 2 ,  ) the likelihood equations are given by (2.4) and (2.5). Now we have  * S S n n    +  'x x      (2.11) From (2.11), we have     1 1 'SI x x n             (2.12) where    1 2 ,        is the solution of (2.4) and (2.5) after replacing ij  by  ij .
  • 4. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 32 Since 1  and 2  have to satisfy these equations, from (2.12) we get the equations  11 12 11 1 211 ' ' / ' /S n S n      (2.13)  21 22 11 21 / /O S n S n   (2.14) From above equations it follows that    11 12 212211 1 11 1 ' ' S n                           =  1 11 1 '    (2.15) Hence, from (2.7), we have   111 1 'x S R  n (2.16) Pre multiplying (2.16) by R we get  1 Rx r =  11 'RS R n (2.17) provided the constraints R = r is to be satisfied. Thus we have    11 1 1 'n RS R Rx r    (2.18)    11 1 1 111 1 ' 'x S R RS R Rx r      (2.19) The ML estimator of  is    122 21 2 12 1 x S S x         3. ML Estimators of ∑ : According to Mardia et al (1989), the likelihood function of p-variate normal distribution with constraints 0 k   , where 0  is known, is  1 1 0 2 , , log log 2n l x k p k k         (3.1) where    1 1 0 0 0 0 'tr S x x          is independent of k. Our problem is to obtain estimate of k for the constraint 11011  k , where 110 is known and
  • 5. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 33 11 : r x r is a submatrix of  . If we let        2221 12110 0    k , then 1 1 0 1 1 0 2 2 2 1 1 1 0 1 2 2 2 0 1 s js r j j k C k                      (3.2) where   11 2212 1 11021   jj trC Further, by considering 1 0   as 1 1 1 1 110.2 110 12 22.10 0 1 1 1 22 21 110.2 22.10 1 k                         where 21 1 22121102.110    k , 1 22.10 22 21 110 12 1 k         ; we get    1 1 1 1 1 1 1 1 1 2 0 22 22 110 11 110 12 22 21 22 21 110 12 22 21 110 22 1 tr S tr S tr S S S S O k k                             (3.3) Since,  1 1 2 110.2 110 1 O k k        (3.4) and  1 1 1 1 2 22.10 22 22 21 110 1 O k k              , (3.5) We have    1 0'x x                      1 1 1 1 21 22 11 22 21 110 1 22 2 1 1 2 1 21 22 2 1 ' ' ' 2kx x x x x x x O k                         (3.6) Using results (3.2), (3.3) and (3.6) in (3.1) we get  l =      1 212 , , . log *n l x k Const p r k O k k          (3.7) where 1 2212 1 1102122110 log2loglog2log   pConst ,    1 1 22 22 222 22 21 * 's b tr S x x             and     11 1 1 21 110 12 221 1 11 1 21 110 12 22 ss s s trc b c                 (3.8) Thus, if  is known, then the mle of k is * ˆk p r    (3.9)
  • 6. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 34 If  is unknown and unconstrained, then the mle's of  is x and that of k is rp *  , together these gives 1 1 22 22ˆ sb tr S k p r       (3.10) Note that for the constraint 22022  k , where 220 is known, the mle's of kˆ for known  is  sp  and that of for unknown  is  sp *  where    1 1 1 11 1111 1 11 1 'rd tr S x x           , * 1 1 11 11r d tr S      with  1 1121 1 22121 1 1121 1 22121       rr trd 4. Illustrative Example : 1. Estimation of sub mean vector with constraint 1 R r  . For 47 female cats the body weights (kgs), heart weights (gms), lungs weights (gms) and Kidney weights (gms) were recorded. The sample mean vector and covariance matrix are 1 2 3 4 x x x x x                1 2 2.36 9.20 3.56 2.76 x x                    0.0735 0.1937 0.2156 0.3072 0.1937 1.8040 0.1969 0.1057 0.2156 0.1969 2.1420 0.1525 0.3072 0.1057 0.1525 2.0136 11 12 21 22 S S S S S                    Note that here x and S are unconstrained mle's of  and . However, from other information we know that  1 2 3 4                      = 1 2           =               5.2 5.3 0.10 5.2 and 11 12 21 22 0.07810 0.15620 0.21745 0.31033 0.15620 1.56200 0.17767 0.07322 0.21745 0.17767 2.15436 0.17338 0.31033 0.07322 0.17338 2.04885                      With above given information we estimate sub-mean vector 1  = 21 under the constraint 1R r  where R : 2 x 2 and r : 2 x 1 are pre-specified as follows :
  • 7. A note on estimation of parameters of multinormal distribution with constraints www.theijes.com The IJES Page 35 andR        91875.012500.0 12500.045000.0 2.37500 9.50000 r        For unknown , we have    1 11 1 2.25821 S R' 0.056431 n R Rx r           111 1 2.49993 ' 10.00096 x S R             and    12 1 2122 2 1 S S xx            = 1 112 121 1x S S x           = 3.84642 3.10968       For known , we have 1 2 2.0804 2.4999 3.9219 , , 0.1706 10.0006 3.1205                              2. Estimation of sub-matrix 11 of  with constraint 11 = k110. For 110 = 0.1 0.2 0.2 2.0       , we have for known  k  = 0.49426 and for unknown , we have k   = 0.4886 REFERENCES [1]. Anderson , T.W. (2003) “An introduction to multivariate Statistical Analysis”. (3rd edition), Wiley Series in probability and Statistics. [2]. Mardia, K.V., Kent, J.T. and Bibby, J.M. (1989). “Multivariate Analysis”. Academic Press, London. [3]. Nagnur, B.N. (2003) : A short note on estimating the mean vector of a multivariate non-singular normal distribution. ISPS Vol-7 83-86.