You are on page 1of 17

Mathematical Theory and Modeling

ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

Asymptotic Properties of Bayes Factor in One- Way Repeated
Measurements Model
Ameera Jaber Mohaisen and Khawla Abdul Razzaq Swadi
Mathematics Department College of Education for Pure Science AL-Basrah University-Iraq
E-mail: Khawla.swade@gmail.com

Abstract
In this paper, we consider the linear one- way repeated measurements model which has only one within units
factor and one between units factor incorporating univariate random effects as well as the experimental error
term. In this model, we investigate the consistency property of Bayes factor for testing the fixed effects linear
one- way repeated measurements model against the mixed one- way repeated measurements alternative model.
Under some conditions on the prior and design matrix, we identify the analytic form of the Bayes factor and
show that the Bayes factor is consistent.

Keywords: One- Way Repeated Measurements Model, ANOVA, Mixed model, Prior Distribution, Posterior
Distribution, covariance matrix, Bayes Factor, Consistent.

1. Introduction
Repeated measurements occur frequently in observational studies which are longitudinal in nature, and in
experimental studies incorporating repeated measures designs. For longitudinal studies, the underlying
metameter for the occasions at which measurements are taken is usually time. Here, interest often centers around
modeling the response as linear or nonlinear function of time, Repeated measures designs, on the other hand,
entail one or more response variables being measured repeatedly on each individual over arrange of conditions.
Here the metameter may be time or it may be a set of experimental conditions (e.g. ,dose levels of a drug).
Repeated measurements analysis is widely used in many fields, for example , in the health and life science,
epidemiology, biomedical, agricultural, industrial, psychological, educational researches and so on.[2],[3],[17]
A balanced repeated measurements design means that the p occasions of measurements are the same for all of
the experimental units. A complete repeated measurements design means that measurements are available each
time point for each experimental unit. A typical repeated measurements design consist of experimental units or
subjects randomized to a treatment (a between-units factor) and remaining on the assigned treatment throughout
the course of the experiment. [2],[3],[17]
A one –way repeated measurements analysis of variance refers to the situation with only one within-units
factor and a multi-way repeated measurements analysis of variance to the situation with more than one within
units factor. This is because the number of within factors, and not the number or between factor, dictates the
complexity of repeated measurements analysis of variance.[2],[3],[4],[12],[17]
In this paper, we consider the linear one- way repeated measurements model which has only one within units
factor and one between units factor incorporating univariate random effects as well as the experimental error
term, we can represent repeated measurements model as a mixed model.
The asymptotic properties of the Bayes factor have been studied mainly in nonparametric density estimation
problems related to goodness of fit testing [1],[5],[7],[8],[9],[15],[16]. Choi, Taeryon and et al in (2009) [6]
studied the semiparametric additive regression models as the encompassing model with algebraic smoothing and
obtained the closed form of the Bayes factor and studied the asymptotic behavior of the Bayes factor based on
the closed form. Mohaisen, A. J. and Abdulhussain , A. M. in (2013)[11] investigated large sample properties of
the Bayes factor for testing the pure polynomial component of spline null model whose mean function consists
of only the polynomial component against the fully spline semiparametric alternative model whose mean
function comprises both the pure polynomial and the component spline basis functions. In this paper, we
investigate the consistency property of Bayes factor for testing the fixed effects linear one- way repeated
measurements model against the mixed one- way repeated measurements alternative model. Under some
conditions on the prior and design matrix, we identify the analytic form of the Bayes factor and show that the
Bayes factor is consistent.

143

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

2. One- Way Repeated Measurements Model
Consider the model
yijk = μ + τj + δi(j) + γk + (τγ)jk + eijk

(1)

Where
i=1,….,n is an index for experimental unit within group j ,
j=1,…,q is an index for levels of the between-units factor (Group) ,
k=1,…,p is an index for levels of the within-units factor (Time) ,
yijk

is the response measurement at time k for unit i within group j,

µ

is the overall mean ,

τj

is the added effect for treatment group j ,

δ i(j)

is the random effect for due to experimental unit i within treatment group j , 𝛾𝑘

is the added effect for time k ,

(τγ)jk is the added effect for the group j × time k interaction ,
e ijk

is the random error on time k for unit i within group j ,

For the parameterization to be of full rank, we imposed the
following set of conditions
∑qj=1 τj=0

∑pk=1 γk=0

,

∑pk=1(τγ)jk=0

,

∑qj=1(τγ)jk=0

for each k=1,…,p

for each j=1,…,q .

And we assumed that the eijk and δi(j) are indepndent with
eijk

~ i.i.d N (0,σ2
e)

δij

,

(2)

~ i.i.d N (0,σ2
δ)

The model (1) is rewritten as follows 𝑌
= 𝑋𝛽 + 𝑍𝑏 + 𝜖

(3)

Where
y111
y112
Y=[ ⋮ ]
ynqp

nqp×1

δ1(1)
δ
b = δ = 1(2)

δ
[ n(q) ]

μ1×1
τq×1
, β = [ γp×1 ]
(τγ)qp×1

,
nq×1

(qp+q+p+1)×1

e111
e112 𝜖
=[ ⋮ ]
enqp

1P×1
, Z=( ⋮
0P×1



0P×1
⋮ )
1P×1 nqp×nq

and 𝑋 = (𝑋1𝑇 , 𝑋2𝑇 , 𝑋3𝑇 , 𝑋4𝑇 )𝑇 is an
nqp×1

(𝑛𝑞𝑝 × (1 + 𝑞 + 𝑝 + 𝑞𝑝)) design matrix of fixed effects.

144

,

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

We assume 𝑛𝑞𝑝 < 1 + 𝑞 + 𝑝 + 𝑞𝑝 + 𝑛𝑞, 𝑛𝑞𝑝 + 𝑠 = 1 + 𝑞 + 𝑝 + 𝑞𝑝 + 𝑛𝑞, (𝑠 is integer number and greater
than 1 ), and we assume that the design matrix as the following way: 𝑇 𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

= 𝑛𝑞𝑝𝐼𝑛𝑞𝑝

∀𝑛𝑞𝑝 ≥ 1

(4)

where: 𝑀𝑛𝑞𝑝
= [𝑋 𝑍𝑟 ] and 𝑍𝑟 be the 𝑛𝑞𝑝 × [𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝)] matrix, and let 𝑍 = [𝑍𝑟 𝑍𝑠 ].
We would like to choose between a Bayesian mixed one- way repeated measurements model and it’s one- way
repeated measurements model (the model without random effects) by the criterion of the Bayes factor for two
hypotheses, 𝐻
0 : 𝑌 = 𝑋𝛽 + 𝜖
i . e 𝐻0 : yijk = μ + τj + γk + (τγ)jk + eijk
versus 𝐻
1 : 𝑌 = 𝑋𝛽 + 𝑍𝑏 + 𝜖

i.e 𝐻

1 : yijk = μ + τj + δi(j) + γk + (τγ)jk + eijk .

(5)

We assume 𝛽 and 𝑏 are independent and the prior distribution for the parameters are 𝛽
~ 𝑁(0, ∑0 ) , where 𝜎𝜇
2
01×𝑞
01×𝑝
0
1×𝑞𝑝

∑0 = 𝜎𝛽2 𝐼1+𝑞+𝑝+𝑞𝑝 =
[
and 𝑏 ~ 𝑁(0, ∑1 )

0𝑞×1 𝜎𝜏

2 𝐼𝑞×𝑞

0𝑞×𝑝

0𝑝×1
0𝑞𝑝×1

0𝑞×𝑝
0𝑞𝑝×𝑞 𝜎𝛾

2 𝐼𝑝×𝑝

, where 𝜎𝛿
211

∑1 = 𝜎𝑏2 𝐼𝑛𝑞 = 𝜎𝛿2 𝐼𝑛𝑞 =

0

[ 0

0 𝜎𝛿
212

0

0𝑞×𝑞𝑝
0𝑝×𝑞𝑝

,

(6)

2 𝜎
(𝜏𝛾) 𝐼𝑞𝑝
×𝑞𝑝 ]

0𝑞𝑝×𝑝

(1+𝑞+𝑝+𝑞𝑝)×(1+𝑞+𝑝+𝑞𝑝)

0

⋯ 0


⋯ 𝜎𝛿2𝑛𝑞 ]

.

(7) 𝑛𝑞

×𝑛𝑞

2.Posterior distribution
From the model (1), we have the following: 𝑌𝑛𝑞𝑝
|𝜉𝑛𝑞𝑝 ~ 𝑁𝑛𝑞𝑝 (𝜉𝑛𝑞𝑝 , 𝜎𝜖2 𝐼𝑛𝑞𝑝 ) , 𝜉𝑛𝑞𝑝 ~ 𝑁𝑛𝑞𝑝 (0𝑛𝑞𝑝 , 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )
where 𝜉𝑖𝑗𝑘
= μ + τj + δi(j) + γk + (τγ)jk + eijk ,for, 𝑖 = 1,2, … , 𝑛, 𝑗 = 1, … , 𝑞 𝑎𝑛𝑑 𝑘 = 1, … , 𝑝.

(8)

Then, the posterior distribution 𝜋1 (𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 ) is the multivariate normal with 𝐸
(𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 , 𝜎𝜖2 ) = ∑2 (∑2 + 𝜎𝜖2 𝐼𝑁 )−1 𝑌𝑛𝑞𝑝 , where N=1+q+p+qp+nq and
2
−1 −1 𝑉𝑎𝑟
(𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 , 𝜎𝜖2 ) = (∑−1
= ∑2 (∑2 + 𝜎𝜖2 𝐼𝑁 )−1 𝜎𝜖2 𝐼𝑁
2 + (𝜎𝜖 𝐼𝑁 ) )
where 𝜎𝜇
2
01×𝑞
01×𝑝
0
1×𝑞𝑝

∑2 = 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 = 𝑋
[ 𝜎𝛿
211
0

0 𝜎𝛿
212

0𝑞×1 𝜎𝜏

2 𝐼𝑞×𝑞

0𝑞×𝑝

0𝑝×1
0𝑞𝑝×1

0𝑝×𝑞
0𝑞𝑝×𝑞 𝜎𝛾

2 𝐼𝑝×𝑝
0𝑞𝑝×𝑝

0𝑞×𝑞𝑝
0𝑝×𝑞𝑝 𝑋𝑇

+ 𝑍

2 𝜎
(𝜏𝛾) 𝐼𝑞𝑝
×𝑞𝑝 ]

0

⋯ 0 𝑍𝑇



⋯ 𝜎𝛿2𝑛𝑞 ]



0
[ 0
Then, 𝑌𝑛𝑞𝑝 ~ 𝑀𝑉𝑁(0, 𝜎𝜖2 𝐼𝑛𝑞𝑝 + ∑2 ).
(9)
2 𝑇

Hence, the distribution of 𝑌𝑛𝑞𝑝 under 𝐻0 and 𝐻1 are respectively 𝑁𝑛𝑞𝑝 (0, 𝜎𝜖 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 ) and 𝑁𝑛𝑞𝑝
(0, 𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ).
(10)
We investigate the consistency of the Bayes factor, under the one- way repeated measurements model with
design matrix of the random effects (𝑍) to the first 𝑛𝑞𝑝 terms. We get
yijk = μ + τj + δK∗ + γk + (τγ)jk + eijk , K ∗ = 1, … . , (𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝)).The covariance matrix of the
distribution of 𝑌𝑛𝑞𝑝 under 𝐻1 as:

145

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org 𝑇 𝜎𝜖

2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 = 𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑀𝑛𝑞𝑝 𝐷𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝜎𝛿

211

0

where ∑1,𝑟= 𝜎𝑏2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) =

⋯ 0 𝜎𝛿

212

0

0
[

(11)

⋯ 0


⋯ 𝜎𝛿2𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) ]


0 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)×𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)


0𝑟1
= 𝜎𝛿2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) and 𝐷𝑛𝑞𝑝 = [ 0
] , where 0𝑟1 , 01𝑟 are the matrices for zeros element with size
01𝑟 ∑1,𝑟
(1 + 𝑞 + 𝑝 + 𝑞𝑝) × (𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝))
and
(𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝)) × (1 + 𝑞 + 𝑝 + 𝑞𝑝)
,
respectively.
Then, we can rewrite (11) by using 𝐷𝑛𝑞𝑝 as: 𝑀𝑛𝑞𝑝
[ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇

+ 𝐷𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
.

(12)

Then, the covariance matrix of the distribution of 𝑌𝑛𝑞𝑝 under 𝐻0 , can be represented as: 𝑇 𝜎𝜖

2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 0𝑟𝑟 𝑍𝑟𝑇 = 𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑀𝑛𝑞𝑝 𝐶𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
, where:

0𝑟1 𝐶𝑛𝑞𝑝
= [ 0
]
01𝑟 0𝑟𝑟
and 0𝑟𝑟 is the matrix for zeros element with size (𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝)) × (𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝)), then
we can rewrite the covariance matrix of the distribution of 𝑌𝑛𝑞𝑝 under 𝐻0 , the following: 𝑀𝑛𝑞𝑝
[ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇

+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝

(13)

and the covariance matrix of the distribution of 𝑌𝑛𝑞𝑝 under 𝐻1 , can be represented as: 𝜎𝜖
2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 = 𝜎𝜖2 𝐼𝑛𝑞𝑝 + [𝑋 𝑍]𝐺𝑛𝑞𝑝 [𝑋 𝑍]𝑇 , where:

0𝑟1 𝐺𝑛𝑞𝑝
= [ 0
]
01𝑟 ∑1
then we can rewrite the covariance matrix of the distribution of 𝑌𝑛𝑞𝑝 under 𝐻1 , the following: 𝜎𝜖
2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

[𝑋 𝑍] [

+ 𝐺𝑛𝑞𝑝 ] [𝑋 𝑍]𝑇

(14)

Lemma(1):Let the covariance matrices of the distribution of 𝑌𝑛𝑞𝑝 under 𝐻𝑂 and 𝐻1 be defined as in (12) and
(13) respectively, and suppose 𝑀𝑛𝑞𝑝 satisfy (4). Then, the following hold. 𝑇 𝑇

1 − 𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
= 𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
= 𝑛𝑞𝑝𝐼𝑛𝑞𝑝
2 − det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) = 𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜇2 )( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜏2 )𝑞 ( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛾2 )𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
)𝑞𝑝

(𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
3 − det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) =
(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜇2 ) ( 𝜎𝜖

2 𝑛𝑞𝑝 𝑞

+ 𝜎𝜏2 ) (

4 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋𝛴0 𝑋 𝑇 )−1 =

1 𝑛𝑞𝑝
2 𝜎𝜖

2 𝑛𝑞𝑝 𝑝

+ 𝜎𝛾2 ) ( 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
)
−1 𝜎𝜖

2 𝑛𝑞𝑝

5 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 = 𝜎𝜖

2 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ]

1 𝑛𝑞𝑝
2 𝑀𝑛𝑞𝑝

[ 𝜎𝜖

2 𝑛𝑞𝑝 𝑞𝑝

( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛿2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑇 𝑀𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝

Proof:𝑇
1 − By multiplying 𝑀𝑛𝑞𝑝 and 𝑀𝑛𝑞𝑝
to equation (4) from right and left side, we have:𝑇 𝑇 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 = 𝑀𝑛𝑞𝑝 𝑛𝑞𝑝𝐼𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝑇

= 𝑀𝑛𝑞𝑝 𝑛𝑞𝑝𝑀𝑛𝑞𝑝 𝑇

= 𝑛𝑞𝑝𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝑇

Multiplying (𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
)−1 to the above equation from left side, we get:𝑇 𝑇 𝑇 𝑇 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
𝑀𝑛𝑞𝑝
(𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
)−1 = 𝑛𝑞𝑝𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
(𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝
)−1 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 = 𝑛𝑞𝑝𝐼𝑛𝑞𝑝 𝜎𝜖
2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

2 − 𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 = 𝑀𝑛𝑞𝑝 [ 𝑇

+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
, and by using (4), we get

146

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014
( 𝜎𝜖

2

+ 𝜎𝜇2 ) 𝑛𝑞𝑝

01×𝑞

0𝑞×1
= [𝑋 , 𝑍𝑟 ]

( 𝜎𝜖

2 𝐼𝑞

+ 𝜎𝜏2 𝐼𝑞 ) 𝑛𝑞𝑝

0𝑝×1

[

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1

0𝑛𝑞×𝑞 𝑛𝑞𝑝

( 𝑛𝑞𝑝

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑞𝑝

0𝑝×𝑛𝑞 𝜎𝜖

2 𝐼𝑝 𝑛𝑞𝑝

+ 𝜎𝛾2 𝐼𝑝 )
( 𝜎𝜖

2 𝐼𝑞𝑝

2
+ 𝜎(𝜏𝛾) 𝐼𝑞𝑝
) 𝑛𝑞𝑝

0𝑛𝑞×𝑝 𝑛𝑞𝑝

( 𝑛𝑞𝑝

+ 𝜎𝜏2 𝐼𝑞 )

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1

0𝑛𝑞×𝑞 𝑛𝑞𝑝

(

[𝑋 , 𝑍𝑟 ]𝑇

0𝑞𝑝×𝑛𝑞 𝜎𝜖
2

0𝑛𝑞×𝑞𝑝 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) ]

01×𝑝

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑞𝑝

0𝑝×𝑛𝑞

01×𝑞 𝜎𝜖
2 𝐼𝑞

0𝑝×1

[

01×𝑝

0𝑞𝑝×𝑝

+ 𝜎𝜇2 )

0𝑞×1
=

(

0𝑝×𝑞 𝜎𝜖

2

www.iiste.org 𝜎𝜖

2 𝐼𝑝 𝑛𝑞𝑝

+ 𝜎𝛾2 𝐼𝑝 ) 𝜎𝜖

2 𝐼𝑞𝑝

0𝑞𝑝×𝑝 𝑛𝑞𝑝

(

0𝑛𝑞×𝑝

2
+ 𝜎(𝜏𝛾) 𝐼𝑞𝑝
) 𝑛𝑞𝑝

0𝑞𝑝×𝑛𝑞 𝜎𝜖
2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) ]

0𝑛𝑞×𝑞𝑝

Then
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) = 𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

(𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

3 − 𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 = 𝑀𝑛𝑞𝑝 [
( 𝜎𝜖

2
+ 𝜎𝜇2 ) 𝑛𝑞𝑝

0𝑞×1

= [𝑋 𝑍𝑟 ]

[

( 𝑛𝑞𝑝

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1

0𝑞×1
= 𝑛𝑞𝑝

0𝑞𝑝×𝑝 𝑛𝑞𝑝

+ 𝜎𝜏2 𝐼𝑞 ) 𝑛𝑞𝑝
(

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1

( 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
)𝑞𝑝

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑞𝑝

0𝑝×𝑛𝑞

[𝑋 𝑍𝑟 ]𝑇

0𝑞𝑝×𝑛𝑞

0𝑛𝑞×𝑞𝑝

( 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) + 𝜎𝛿2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) )

01×𝑝

01×𝑞

01×𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞 𝜎𝜖

2 𝐼𝑝
+ 𝜎𝛾2 𝐼𝑝 ) 𝑛𝑞𝑝

0𝑝×𝑞𝑝

0𝑞𝑝×𝑝

0𝑛𝑞×𝑞 𝜎𝜖

2

+ 𝜎𝛾2 )𝑝 ( 𝜎𝜖

2 𝐼𝑞𝑝
2
+ 𝜎(𝜏𝛾) 𝐼𝑞𝑝
) 𝑛𝑞𝑝

0𝑛𝑞×𝑝 𝜎𝜖

2 𝐼𝑞 𝜎𝜖

2 𝑛𝑞𝑝 𝑇

+ 𝐷𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝

+ 𝜎𝛾2 𝐼𝑝 )

01×𝑞 𝑛𝑞𝑝

(

0𝑝×1

[

( 𝜎𝜖

2 𝐼𝑝

0𝑛𝑞×𝑞 𝜎𝜖

2 𝑛𝑞𝑝
(
+ 𝜎𝜇2 ) 𝑛𝑞𝑝 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

+ 𝜎𝜏2 )𝑞 (

0𝑞×𝑝

+ 𝜎𝜏2 𝐼𝑞 )

0𝑝×1 𝜎𝜖

2 𝑛𝑞𝑝

01×𝑝

01×𝑞 𝜎𝜖
2 𝐼𝑞

+ 𝜎𝜇2 )( 𝑛𝑞𝑝

( 𝜎𝜖

2 𝐼𝑞𝑝 𝑛𝑞𝑝

0𝑛𝑞×𝑝

]

.

0𝑝×𝑛𝑞

2
+ 𝜎(𝜏𝛾) 𝐼𝑞𝑝
)

0𝑛𝑞×𝑞𝑝

0𝑞𝑝×𝑛𝑞 𝑛𝑞𝑝
( 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) + 𝜎𝛿2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) )

Then

4 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 )−1 = {𝑀𝑛𝑞𝑝 [ 𝑇

= 𝑀𝑛𝑞𝑝

=
=
=

−1 𝐼𝑛𝑞𝑝 𝑛𝑞𝑝

−1

(𝑛𝑞𝑝)2

[ 𝜎𝜖

2 𝐼

+ 𝐶𝑛𝑞𝑝 ] 𝑛𝑞𝑝
𝑛𝑞𝑝
−1

[ 𝜎𝜖

2 𝑛𝑞𝑝 𝑇 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

1
(𝑛𝑞𝑝)2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝
−1 𝑇 𝑛𝑞𝑝𝐼𝑛𝑞𝑝

𝑀𝑛𝑞𝑝

1 𝜎𝜖

2 𝑀𝑛𝑞𝑝

[ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

−1

2

+ 𝜎𝜏2 )𝑞 ( 𝜎𝜖

2 𝑛𝑞𝑝 𝑇

+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
}−1 𝑀𝑛𝑞𝑝

−1
−1 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ]
[

2 𝜎𝜖 𝜎

+ 𝜎𝜇2 )( 𝜖 𝑛𝑞𝑝 𝑛𝑞𝑝 𝜎𝛿

2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) = (𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝
−1

+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝

−1 𝑛𝑞𝑝𝐼𝑛𝑞𝑝
−1

+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝

−1

−1 𝐼𝑛𝑞𝑝 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

(by lemma 1)

5 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 = {𝑀𝑛𝑞𝑝 [ 𝜎𝜖

2 𝑇 𝐼

+ 𝐷𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
}−1 𝑛𝑞𝑝
𝑛𝑞𝑝

147

+ 𝜎𝛾2 )𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
)𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

+

]

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014 𝑇

= 𝑀𝑛𝑞𝑝

=
=
=

−1 𝐼𝑛𝑞𝑝

−1

[ 𝜎𝜖

2 𝐼

+ 𝐷𝑛𝑞𝑝 ] 𝑛𝑞𝑝
𝑛𝑞𝑝 𝑇 𝑛𝑞𝑝𝐼𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑛𝑞𝑝

1

(𝑛𝑞𝑝)2

−1

1
(𝑛𝑞𝑝)2

−1 𝑀𝑛𝑞𝑝

−1
−1 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

[ 𝑇 𝑇 𝑀𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

−1

[

www.iiste.org

+ 𝐷𝑛𝑞𝑝 ]

−1 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝
−1

−1 𝐼𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

−1 𝑛𝑞𝑝𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

−1 𝑀𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

. ∎

+ 𝐷𝑛𝑞𝑝 ]

3- The Bayes factor
The Bayes factor for testing problem (5) is given by:
m(yijk |𝐻𝑜 ) 𝐵
01 (yijk ) =
m(yijk |𝐻1 )
1

=

2 𝜎
2 𝜎
2 𝜎
2 𝑛𝑞𝑝
−(1+𝑞+𝑝+𝑞)
√2𝜋𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖 +𝜎2
)( 𝜖 +𝜎2)𝑞 ( 𝜖 +𝜎2
)𝑝 ( 𝜖 +𝜎2
)𝑞𝑝 (𝜎2 𝜖
) 𝑛𝑞𝑝
𝜇 𝑛𝑞𝑝 𝜏 𝑛𝑞𝑝
𝛾 𝑛𝑞𝑝
(𝜏𝛾)

−1 𝜎
2
1 𝑇
)𝑌 𝑀

[ 𝜖𝐼
+ 𝐶𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝 𝑛𝑞𝑝
} 𝑛𝑞𝑝
2 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

1
2 𝑇

(
exp{− 𝑌𝑛𝑞𝑝

1
1 𝑇
1
exp{− 𝑌𝑛𝑞𝑝
(
[𝑋 𝑞 𝑝 𝑞𝑝

2
2 𝑛𝑞𝑝
2
2
2
2
2
√2𝜋(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖 +𝜎2 )( 𝜎𝜖 +𝜎2 ) ( 𝜎𝜖 +𝜎2 ) ( 𝜎𝜖 +𝜎2 ) ( 𝜎𝜖 +𝜎2 )𝑛𝑞 𝜇 𝜏 𝛾 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

(𝜏𝛾) 𝑛𝑞𝑝
𝛿

2

= 𝑞

2 𝑝

2

2

2 )( 𝜎𝜖 +𝜎 2 ) ( 𝜎𝜖 +𝜎 2 ) ( 𝜎𝜖 +𝜎 2 )
√(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑞𝑝 𝑛𝑞𝑝 𝜎

2 𝜖𝐼 𝑍

][𝑛𝑞𝑝 𝑛𝑞𝑝
+ 𝐺𝑛𝑞𝑝 ] 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

−1

+ 𝐶𝑛𝑞𝑝 ] 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

−[𝑋 𝑍] [

[𝑋 𝑍]𝑇 )𝑌𝑛𝑞𝑝}

2 𝜎

( 𝜖 +𝜎𝛿2 )𝑛𝑞 𝑛𝑞𝑝

2
2
2
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 (𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾 𝜖

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

(𝑀𝑛𝑞𝑝 [

−1

−1

+ 𝐺𝑛𝑞𝑝 ]

exp{ −

1 𝑌𝑇

2𝑛𝑞𝑝2 𝑛𝑞𝑝

[𝑋 𝑍]𝑇 )𝑌𝑛𝑞𝑝 }.

(15)

Now let the approximation of 𝐵01 with design matrix of the random effects (𝑍) to the first 𝑛𝑞𝑝 terms as
following
2 𝐵

̃01 =

2

2

2

2

2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 ( 𝜎𝜖 +𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝛿 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

2
2
2
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 (𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾 𝜖

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝜎𝜖

2

(𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝

−𝑀𝑛𝑞𝑝 [

−1 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

exp{−

1 𝑌𝑇

2𝑛𝑞𝑝2 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

)𝑌𝑛𝑞𝑝 }.

(16)

Then,
2

log𝐵̃01 =

1
2 𝜎𝜖

2

(𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

log

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝜎𝜖

2 𝑛𝑞𝑝 𝑛𝑞𝑝

2 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

−𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝜎

(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜖 +𝜎𝛿2 ) 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

1 𝑇 𝑌𝑛𝑞𝑝

(𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

2 𝜎𝜖

2 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

)𝑌𝑛𝑞𝑝
2 𝜎

(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜖 +𝜎𝛿2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝 𝑛𝑞𝑝

1+𝑞+𝑝+𝑞𝑝 (𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

1 𝑇

2 {𝑌𝑛𝑞𝑝 𝑀𝑛𝑞𝑝 {[

(𝑛𝑞𝑝)

−1 𝑇 𝐷𝑛𝑞𝑝

] } 𝑀𝑛𝑞𝑝 𝑌𝑛𝑞𝑝
}
2

1 𝑌𝑇

2𝑛𝑞𝑝2 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

)𝑌𝑛𝑞𝑝 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝

1+𝑞+𝑝+𝑞𝑝 (𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

→ 2 log𝐵̃01 = log

= log

− 𝑛𝑞𝑝

−1 𝜎𝜖

2 𝑛𝑞𝑝

2

−𝑀𝑛𝑞𝑝 [

2

2
2
2 𝜎
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 (𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝
1+𝑞+𝑝+𝑞𝑝 ( 𝜖 +𝜎𝜇 𝜏 𝛾 𝜖

(𝜏𝛾)

−1

2 log𝐵̃01 = log

2 𝜎

2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 ( 𝜎𝜖 +𝜎 2 )
(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝛿 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝜎

(𝑛𝑞𝑝)𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) ( 𝜖 +𝜎𝛿2 ) 𝑛𝑞𝑝 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)

(𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

1 𝑇

− (𝑛𝑞𝑝)2 {𝑌𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

148 𝜎𝜖

2 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

− [

+

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014
0

01×𝑞

0𝑞×1
0𝑝×1

0𝑞×𝑞
0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

[0𝑛𝑞×1

01×𝑝

0𝑛𝑞×𝑞

= log(

01×𝑞𝑝

www.iiste.org

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑝
0𝑞𝑝×𝑝

0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝 𝜎𝜖

2 +𝑛𝑞𝑝𝜎𝛿2 𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝜎𝜖
2

)

( 𝑇 𝑀𝑛𝑞𝑝 𝑌𝑛𝑞𝑝

}

(𝑛𝑞𝑝)2 𝜎𝛿2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

]𝑛𝑞𝑝×𝑛𝑞𝑝 𝑇

− {𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 }

where 𝑄𝑛𝑞𝑝

=

1
(𝑛𝑞𝑝)2

01×𝑝

0

01×𝑞

0𝑞×1
0𝑝×1

0𝑞×𝑞
0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

[0𝑛𝑞×1

0𝑛𝑞×𝑞 𝑀𝑛𝑞𝑝

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑝
0𝑞𝑝×𝑝

0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

( 𝑇 𝑀𝑛𝑞𝑝

.

(𝑛𝑞𝑝)2 𝜎𝛿2

(𝜎𝜖2 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

]𝑛𝑞𝑝×𝑛𝑞𝑝

The Bayes factor for testing the model in (5) satisfy is consistent if [6] 𝑙𝑖𝑚𝑛𝑞𝑝
→∞ 𝐵01 = ∞
, 𝑎𝑙𝑚𝑜𝑠𝑡
𝑠𝑢𝑟𝑒𝑙𝑦 𝑖𝑛 𝑝𝑟𝑜𝑏𝑎𝑏𝑙𝑖𝑡𝑦 𝑃0∞ under 𝐻0 , 𝑙𝑖𝑚𝑛𝑞𝑝
→∞ 𝐵01 = 0
, 𝑎𝑙𝑚𝑜𝑠𝑡 𝑠𝑢𝑟𝑒𝑙𝑦 𝑖𝑛 𝑝𝑟𝑜𝑏𝑎𝑏𝑙𝑖𝑡𝑦 𝑃1∞ under 𝐻1 ,

(17)
(18)

Lemma(2):let 𝑌𝑛𝑞𝑝 = (𝑦111 , … , 𝑦𝑛𝑞𝑝 )𝑇 where 𝑦𝑖𝑗𝑘 ′𝑠 are independent normal random variable
with mean μ + τj + γk + (τγ)jk and variance 𝜎𝜖2 = 1, then there exist a positive constant 𝐶1 such that
1 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝑆𝑛𝑞𝑝

,1

[∑𝑖=1 𝑛𝑞𝑝

→∞

→ 𝑇

log(1 + 𝑛𝑞𝑝𝜎𝛿2 ) − 𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ] > 𝐶1 , with probability tending to 1, i.e. log𝐵̃01
nqpσ2
δ

nqp

∞, in probability, where Snqp,1 = ∑i=1

.

(1+nqpσ2
δ)

Proof:
Let Xnqp = ( X111 , X121 , … , Xnqp ) are independent standard normal random variables, Note that Ynqp =d Xnqp + 𝑋𝛽
and
T
Ynqp
Q nqp Ynqp = (Xnqp + 𝑋𝛽)T Q nqp (X nqp + 𝑋𝛽)
T
T
= Xnqp
Q nqp Xnqp + (𝑋𝛽)T Q nqp (𝑋𝛽) + Xnqp
Q nqp (𝑋𝛽) + (𝑋𝛽)T Q nqp Xnqp .
By the property of the design matrix (4) get
01×𝑞

01×𝑝

01×𝑞𝑝

01×𝑛𝑞

0𝑞×1
0𝑝×1

0𝑞×𝑞
0𝑝×𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑝×𝑝
0𝑞𝑝×𝑝

0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

[0𝑛𝑞×1

0𝑛𝑞×𝑞

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

0

(𝑋𝛽)T

Q nqp =

(𝑋𝛽)T

1
(nqp)2

Mnqp

T
Mnqp
=

1
(𝑛𝑞𝑝)2

0
0𝑞×1

01×𝑞
0𝑞×𝑞

[𝜇, 𝜏, 𝛾, (𝜏𝛾)]𝑛𝑞𝑝 (𝐼1+𝑞+𝑝+𝑞𝑝 , 0𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝 ) 0𝑝×1
0𝑞𝑝×1

0𝑝×𝑞

0𝑛𝑞×1
[

0𝑞𝑝×𝑞
0𝑛𝑞×𝑞

(

(𝑛𝑞𝑝)2 𝜎𝛿2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

01×𝑝
0𝑞×𝑝
0𝑝×𝑝
0𝑞𝑝×𝑝

01×𝑞𝑝
0𝑞×𝑞𝑝
0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

(

2

(𝑛𝑞𝑝) 𝜎2𝛿
2

nqpσ2
δ

(1+nqpσ2
δ)

= Snqp,1

Note that

149

2

(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿 )𝜎2𝜖

T
T
Then Ynqp
Q nqp Ynqp = Xnqp
Q nqp Xnqp .
Now by using the expectation formula of the quadratic form[4],[10],[12],[14], get
nqp

]

01×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞 𝑛𝑞𝑝

(𝐼1+𝑞+𝑝+𝑞𝑝 , 0𝑛𝑞−(1+𝑞+𝑝+𝑞𝑝 ) [𝜇, 𝜏, 𝛾, (𝜏𝛾)]𝑇 = 0.

T
0 < 𝐸(Xnqp
Q nqp Xnqp ) = tr(Q nqp ) = ∑i=1

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
]

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

Q2nqp

1

=

(nqp)4

Mnqp

0

01×𝑞

0𝑞×1
0𝑝×1

0𝑞×𝑞
0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

[0𝑛𝑞×1
0
0𝑞×1

01×𝑞
0𝑞×𝑞

0𝑝×1

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1
[

0𝑛𝑞×𝑞

1

=

(nqp)3

Mnqp

01×𝑝

0𝑛𝑞×𝑞

01×𝑝
0𝑞×𝑝
0𝑝×𝑝
0𝑞𝑝×𝑝

01×𝑞𝑝
0𝑞×𝑞𝑝
0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

(

01×𝑞

01×𝑝

0𝑞×1
0𝑝×1

0𝑞×𝑞
0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞
0𝑛𝑞×𝑞

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑝
0𝑞𝑝×𝑝

0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

(

01×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

0

[0𝑛𝑞×1

www.iiste.org

(𝑛𝑞𝑝)2 𝜎𝛿2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

]

MTnqp

2

(𝑛𝑞𝑝) 𝜎2𝛿
2

2

T
Mnqp
Mnqp

(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿 )𝜎2𝜖

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
]

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

0𝑝×𝑝
0𝑞𝑝×𝑝

0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

(

T
Mnqp

(𝑛𝑞𝑝)2 𝜎𝛿2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

)2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

]

and
1

tr(Q2nqp ) =

(nqp)2

∑nqp
i=1 (

(nqp)2 σ2
δ 2
)
1+nqpσ2
δ

nqpσ2
δ

nqp
= ∑i=1 (

1+nqpσ2
δ

)2 = nqp3 (

σ2
δ

1+nqpσ2
δ

)2 = Snqp,2 .

Using the variance formula of the quadratic form[4],[10],[12],[14], get
nqp

T
var(Xnqp
Q nqp Xnqp ) = 2tr(Q2nqp ) = 2Snqp,2 = 2 ∑i=1 (
Snqp,3 −Snqp,1

Let cnqp =
Pr {

1

2Snqp,1

)2 . 𝑛𝑞𝑝

, where Snqp,3 = ∑𝑖=1 log(1 + 𝑛𝑞𝑝𝜎𝛿2 ) = 𝑛𝑞𝑝 log(1 + 𝑛𝑞𝑝𝜎𝛿2 ) . Then,

nqp−(1+q+p+qp)

[∑i=1

Snqp,1

nqpσ2
δ

1+nqpσ2
δ

T
log(1 + nqpσ2δ )2 − Ynqp
Q nqp Ynqp ] ≤ cnqp } = Pr [

{YT
nqp Qnqp Ynqp }
Snqp,1

Snqp,3
Snqp,1

cnqp ]
= Pr [
= Pr [

{YT
nqp Qnqp Ynqp }

Snqp,1
1

E(YT
nqp Qnqp Ynqp )
Snqp,1

var(YT
nqp Qnqp Ynqp )

1

Snqp,3
Snqp,1

− cnqp −

T
T
{(Ynqp
Q nqp Ynqp ) − E(Ynqp
Q nqp Ynqp )} ≥

Snqp,1

Snqp,2 −Snqp,1
(
)2
2Snqp,1

S2
nqp,1

nqp→∞

8Snqp,2

=

(Snqp,3 −Snqp,1 )2

E(YT
nqp Qnqp Ynqp )
Snqp,1

]

Snqp,3 −E(YT
nqp Qnqp Ynqp )
Snqp,1

− cnqp ]

0, from the Chebyshev inequality[13] .

Since Snqp,3 − Snqp,1 ≥ c1 Snqp,1 for some c1 > 0 and let cnqp ≥ c1 /2.
Then, there exists a positive constant C1 = c1 /2 such as
1

nqp−(1+q+p+qp)

Snqp,1

[∑i=1

T
log(1 + nqpσ2δ )2 − Ynqp
Q nqp Ynqp ] > C1 , with probability tending to 1.

Then,
nqp→∞
T
̃01 = ∑nqp−(1+q+p+qp)
2logB
log(1 + nqpσ2δ )2 − Ynqp
Q nqp Ynqp →
∞ in probability under H0 i.e.
i=1
nqp→∞

̃01 →
logB

Lemma 3: Suppose we have the one- way repeated measurements model, then
1 𝑡𝑟

(

1 𝑛𝑞𝑝

2 𝑆𝑛𝑞𝑝

,1 𝑀𝑛𝑞𝑝

[ 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

−1 𝑇 𝑀𝑛𝑞𝑝

− [𝑋 𝑍] [ 𝜎𝜖

2 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐺𝑛𝑞𝑝 ] 𝑛𝑞𝑝

→∞

[𝑋 𝑍]𝑇 ) →

0

Proof: 𝜎𝜖
2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

Since[𝑋 𝑍] [ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

+ 𝐺𝑛𝑞𝑝 ] [𝑋 𝑍]𝑇 −𝑀𝑛𝑞𝑝 [ 𝑇

+ 𝐷𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
is the covariance matrix of 𝜉

∗ = 𝛿𝐾∗ = 𝑍𝑠 𝑏, 𝐾 ∗ = 𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 + 𝑞𝑝), … , 𝑛𝑞𝑝 + 𝑠, it is a positive definite matrix. Thus
1 𝑛𝑞𝑝
2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

Thus, 𝑡𝑟 (

1 𝑛𝑞𝑝

2

−1

+ 𝐷𝑛𝑞𝑝 ] 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[ 𝑇 𝑀𝑛𝑞𝑝

+

1

[𝑋 𝑍] [ 𝑛𝑞𝑝

2
−1 𝑇 𝐷𝑛𝑞𝑝

] 𝑀𝑛𝑞𝑝
) 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

≥ 𝑡𝑟(

1 𝑛𝑞𝑝
2

150

+ 𝐺𝑛𝑞𝑝 ]

−1

[𝑋 𝑍]𝑇 is also a positive definite matrix. 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

[𝑋 𝑍] [

−1

+ 𝐺𝑛𝑞𝑝 ]

[𝑋 𝑍]𝑇 ), also

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014 𝑡𝑟

(𝑍𝑠𝑇

1

= ‖𝑍𝑠 ‖ ‖

1

1 𝑛𝑞𝑝

2 { (( 𝜎𝜖

2 𝜎𝜖

2

+ 𝐷𝑛𝑞𝑝 ] 𝑛𝑞𝑝 𝑞 𝜎

2
∑𝑛𝑞𝑝−(1+𝑞=𝑝+𝑞𝑝)
( 𝜖 𝑖
=1 𝑛𝑞𝑝

+ 𝜎

2𝛿 )

−2

−1

−1

+ 𝐷𝑛𝑞𝑝 ] 𝑇

) 𝑀𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

‖ by the Cauchy-Schwarz inequality, where
1 𝜎𝜖

2 𝑝 𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2 𝑛𝑞𝑝

2 𝜎𝜖

2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

1 𝑇 𝑀𝑛𝑞𝑝

) ≤ √𝑡𝑟(𝑍𝑠𝑇 𝑍𝑠 )𝑡𝑟 ( 𝑇 𝑀𝑛𝑞𝑝

‖ = { 𝑡𝑟 (( 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

+ 𝜎𝜇2 ) + ∑𝑖=1 ( 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]
−1 𝜎𝜖

2 𝑛𝑞𝑝

1 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2

=[ 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2

www.iiste.org

+ 𝜎𝜏2 ) + ∑𝑖=1 ( 𝑞𝑝 𝑛𝑞𝑝

1 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇

+ 𝐷𝑛𝑞𝑝 ] 𝑀𝑛𝑞𝑝
)−2 )}2 𝑀𝑛𝑞𝑝

[

+ 𝜎𝛾2 ) + ∑𝑖=1 (

−2 𝜎𝜖

2

2
+ 𝜎(𝜏𝛾)
)) 𝑛𝑞𝑝

+

1
2

}] .

Since 𝜎
2 𝜎

2 𝑞 𝜎

2 𝑝

−2 𝜎

2 𝑞𝑝 𝑛𝑞𝑝

2
((𝜎𝜖2+𝑛𝑞𝑝𝜎𝜇2)+𝑞(𝜎𝜖2+𝑛𝑞𝑝𝜎𝜏2)+𝑝(𝜎𝜖2 +𝑛𝑞𝑝𝜎𝛾2 )+𝑞𝑝(𝜎𝜖2+𝑛𝑞𝑝𝜎2(𝜏𝛾) ))

=

1
((𝜎2𝜇 )+𝑞(𝜎2𝜏 )+𝑝(𝜎2𝛾 )+𝑞𝑝(𝜎2(𝜏𝛾) ))

∑𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
( 𝑖
=1 𝜎𝜖

2

1 𝑛𝑞𝑝

2 𝑀𝑛𝑞𝑝 [

−2

= 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)
2 𝜎

( 𝜖 +𝜎𝛿2 )

−1 𝜎𝜖

2 𝑛𝑞𝑝

= 𝑛𝑞𝑝

2
((𝑛𝑞𝑝𝜎𝜇2)+𝑞(𝑛𝑞𝑝𝜎𝜏2 )+𝑝(𝑛𝑞𝑝𝜎𝛾2)+𝑞𝑝(𝑛𝑞𝑝𝜎2(𝜏𝛾) ))

2

≤ 𝑛𝑞𝑝

Then, ‖

2
2
2
2 𝜎
2 𝜖
+𝜎2 )+𝑞( 𝜎𝜖 +𝜎2 )+𝑝( 𝜎𝜖 +𝜎2 )+𝑞𝑝( 𝜎𝜖 +𝜎2 ))
((𝑛𝑞𝑝 𝜇 𝜏 𝛾

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

2

= 𝑂(1), and

2

+ 𝜎𝛿2 ) 𝑛𝑞𝑝

2

1

= 𝜖

2
+ 𝜎𝜇2 ) + ∑𝑖=1 ( 𝜖 + 𝜎𝜏2 ) + ∑𝑖=1 ( 𝜖 + 𝜎𝛾2 ) + ∑𝑖=1 ( 𝜖 + 𝜎(𝜏𝛾)
))
((𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝑛𝑞𝑝

2 𝜎

( 𝜖 +𝜎𝛿2 )

2

= 𝑛𝑞𝑝

(𝑛𝑞𝑝)3
2
(𝜎𝜖2 +𝑛𝑞𝑝𝜎𝛿2 )

(𝑛𝑞𝑝)3
2

(𝑛𝑞𝑝𝜎𝛿2 )

= 𝑂(𝑛𝑞𝑝).

1 𝑇 𝑀𝑛𝑞𝑝

‖ = 𝑂{(𝑛𝑞𝑝)2 }.

Since 𝑠𝑢𝑝‖𝑍𝑠 ‖ = 1 from the distances between the elements of design matrix 𝑍, then ‖𝑍𝑠 ‖2 ≤ 𝑛𝑞𝑝.
1

Thus ,
1 𝑆𝑛𝑞𝑝
,1 𝑆𝑛𝑞𝑝

,1 𝑡𝑟

( 𝑡𝑟

(

1 𝑀𝑛𝑞𝑝

[ 𝑛𝑞𝑝

2

1 𝑛𝑞𝑝

2 𝜎𝜖

2 𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[ 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

1 𝑆𝑛𝑞𝑝
,1
−1 𝐺𝑛𝑞𝑝

] 𝑛𝑞𝑝𝜎𝛿

2

≤ 𝑛𝑞𝑝𝑆𝑛𝑞𝑝

,1 𝜎𝛿

2 𝑆𝑛𝑞𝑝
,1

1 𝑛𝑞𝑝

−1 𝑇 𝑀𝑛𝑞𝑝


1 𝑇 𝑀𝑛𝑞𝑝

{ 𝑛𝑞𝑝

2

1

[𝑋 𝑍] [ 𝑛𝑞𝑝

2

[𝑋 𝑍] [ 𝜎𝜖

2 𝑛𝑞𝑝 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐺𝑛𝑞𝑝 ]

−1

+ 𝐺𝑛𝑞𝑝 ] [𝑋 𝑍]𝑇 −

[𝑋 𝑍]𝑇 ) =
1 𝑛𝑞𝑝

2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

+

[𝑋 𝑍]𝑇 )

+ 𝐺𝑛𝑞𝑝 ] 𝑇

2
∑𝑛𝑞𝑝+𝑠 𝐾
∗ =𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝜎𝛿 (𝑍𝐾 ∗

[𝑋 𝑍]𝑇 )𝑇 ) ≤

−1

−1 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇 𝐷𝑛𝑞𝑝

] 𝑀𝑛𝑞𝑝
} [𝑋 𝑍] [

= 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

1 𝑛𝑞𝑝
2

2 𝑀𝑛𝑞𝑝 [ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

−1 𝜎𝜖

2 𝑛𝑞𝑝

∑𝑛𝑞𝑝+𝑠 𝐾
∗ =𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 1 ≤

−1

+ 𝐷𝑛𝑞𝑝 ] 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝐾

∗ 𝜎𝛿2 𝑛𝑞𝑝
𝑆𝑛𝑞𝑝,1

= 𝑇 𝑀𝑛𝑞𝑝

(𝑍𝐾𝑇 ∗

1 𝑛𝑞𝑝
2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

[𝑋 𝑍] [

+

2 𝑛𝑞𝑝

+𝑠 𝑇 𝑀𝑛𝑞𝑝

‖ ∑𝐾∗=𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)‖𝑧𝐾∗ ‖2 𝐾

∗ 𝜎𝛿2
(𝑛𝑞𝑝)2 𝜎2 𝛿

(1+𝑛𝑞𝑝𝜎2 𝛿
)

= 𝐾

∗ (1+𝑛𝑞𝑝𝜎𝛿2 )
(𝑛𝑞𝑝)2

Thus
1 𝑡𝑟

( 𝑆𝑛𝑞𝑝

,1 𝑛𝑞𝑝
→∞

1 𝑛𝑞𝑝

2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

−1

+ 𝐷𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝

1 𝑛𝑞𝑝
2

[𝑋 𝑍] [ 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝

+ 𝐺𝑛𝑞𝑝 ]

−1

[𝑋 𝑍]𝑇 ) ≤ 𝐾

∗ (1+𝑛𝑎𝑝𝜎𝛿2 )
(𝑛𝑞𝑝)2

0

Theorem(1): Suppose that the true model is the one- way repeated measurements model with fixed effect 𝑛𝑞𝑝

only (the model under 𝐻0 in 5), and let 𝑃0 denote the true distribution of the data, with probability density
function 𝑃0 ( yijk |μ, τj , γk , (τγ)jk ) = 𝛷{(𝑦𝑖𝑗𝑘 − μ − τj − γk − (τγ)jk )/ 𝜎𝜖 } , where 𝛷(∙) is the standard normal
density, then, the Bayes factor is consistent under the null hypothesis 𝐻0 i.e 𝑛𝑞𝑝

lim𝑛𝑞𝑝→∞ 𝐵01 = ∞ in probability 𝑃0 .

Proof:
1

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

log 𝐵01 = log
1

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )

= log

+

1 𝑇

− 𝑌𝑛𝑞𝑝
( (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝
2

1
2

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

151

1 𝑇

− 𝑌𝑛𝑞𝑝
(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1
2

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

1 𝑇

−(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 )𝑌𝑛𝑞𝑝 − 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 +
2 𝑍

∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝
1
= log𝐵̃01 + log
2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟𝑍𝑟𝑇 )

1 𝑇

− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 +
2 𝑍

∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝
Since (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) is a positive definite matrix, 𝑑𝑒𝑡
(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) ≥ 𝑑𝑒𝑡(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ), thus
1

log

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )

≥0

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) 𝑇 𝐸

( 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝 ) = 𝜎𝜖2 𝑡𝑟((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋

∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 ) + (𝑋𝛽)𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −
(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )(𝑋𝛽) ,
since (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 , (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1
and (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 are all nonnegative definite matrices, it follows that
0 ≤ (𝑋𝛽)𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + (𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )(𝑋𝛽) ≤ (𝑋𝛽)𝑇
(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 (𝑋𝛽).
By the property of design matrix (4), we get 𝑇 𝑋

𝑇 𝑀𝑛𝑞𝑝 = 𝑀𝑛𝑞𝑝 𝑋
= 𝑛𝑞𝑝 (𝐼1+𝑞+𝑝+𝑞𝑝 , 0𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝 ).
Now as the proof of lemma (3), we get:
(𝑋𝛽)𝑇 (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 (𝑋𝛽) =

1
(𝑛𝑞𝑝)2

(𝑋𝛽)𝑇 𝑀𝑛𝑞𝑝 [ 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ]

−1 𝑇 𝑀𝑛𝑞𝑝

(𝑋𝛽)

1
[𝜇, 𝜏, 𝛾, (𝜏𝛾)]𝑛𝑞𝑝 (𝐼1+𝑞+𝑝+𝑞𝑝 , 0𝑛𝑞−(1+𝑞+𝑝+𝑞𝑝 )
(𝑛𝑞𝑝)2

=

( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜇2 )

0𝑞×1

( 𝜎𝜖

2 𝑛𝑞𝑝

0𝑝×1

01×𝑞𝑝

01×𝑛𝑞

0𝑞×𝑝

0𝑞×𝑞𝑝

0𝑞×𝑛𝑞

(

0𝑝×𝑞𝑝

0𝑝×𝑛𝑞 𝜎𝜖

2 𝑛𝑞𝑝

0𝑞𝑝×𝑞

0𝑛𝑞×1

01×𝑝

+ 𝜎𝜏2 )𝐼𝑞
0𝑝×𝑞

0𝑞𝑝×1
[

01×𝑞

+ 𝜎𝛾2 )𝐼𝑝

0𝑞𝑝×𝑝

( 𝑛𝑞𝑝

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞 𝜎𝜖

2

2
+ 𝜎(𝜏𝛾)
)𝐼𝑞𝑝

0𝑛𝑞×𝑞𝑝

−1 𝑛𝑞𝑝

0𝑞𝑝×𝑛𝑞
( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛿2 )𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) ]

(𝐼1+𝑞+𝑝+𝑞𝑝 , 0𝑛𝑞−(1+𝑞+𝑝+𝑞𝑝 ) [𝜇, 𝜏, 𝛾, (𝜏𝛾)]𝑇
= 𝜇2 [ 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜇2 ]

−1 𝑞

+ ∑𝑖=1 𝜏𝑖2 [ 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜏2 ]

−1 𝑝

+ ∑𝑖=1 𝛾𝑖2 [ 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛾2 ]

−1 𝑞𝑝

+ ∑𝑖=1(𝜏𝛾)2𝑖 [ 𝜎𝜖

2 𝑛𝑞𝑝

−1

2
+ 𝜎(𝜏𝛾)
]

= 𝑂(1).
Since 𝑝𝑟 {
1 𝑆𝑛𝑞𝑝
,1

{

1 𝑌
𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 𝑆𝑛𝑞𝑝
,1 𝑛𝑞𝑝

+ 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝 > 𝜀 } ≤ 𝑇

((𝜎 2 𝐼 𝑇 𝑇

−1
2 𝑇 𝑇

−1 𝐸
( 𝑌𝑛𝑞𝑝 𝜖
𝑛𝑞𝑝 +𝑋∑0 𝑋 +𝑍𝑟 ∑1,𝑟 𝑍𝑟 ) −(𝜎𝜖 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 +𝑍∑1 𝑍 ) )𝑌𝑛𝑞𝑝 ) 𝜀

1 𝑡𝑟

((𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )−1 𝑆𝑛𝑞𝑝

,1 𝜀

(𝑋𝛽)𝑇 (𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 (𝑋𝛽) 𝑛𝑞𝑝→∞

+

}

1 𝑆𝑛𝑞𝑝
,1

0 (by the Markov inequality[13], where 𝜀 > 0) . 𝜀

Since (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 is nonnegative definite 𝑇 𝑌𝑛𝑞𝑝

((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝 is nonnegative .

Therefore,
1 𝑇
− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝 = 𝑜(𝑆𝑛𝑞𝑝,1 )
2
1 𝑇
− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )−1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 )−1 )𝑌𝑛𝑞𝑝
2
⇒ 𝑆𝑛𝑞𝑝
,1
= 𝑜(1)
Then, as in lemma (2) there exists a constant 𝐶1 such that

152

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014
1 𝑆𝑛𝑞𝑝
,1

1

log 𝐵01 ≥ 𝑆𝑛𝑞𝑝

,1

www.iiste.org

log 𝐵̃01 + 𝑜(1) > 𝐶1 + 𝑜(1).

That is, 𝑛𝑞𝑝
→∞ 𝐵

01 ≥ exp( 𝑆𝑛𝑞𝑝,1 { 𝐶1 + 𝑜(1)}) → 𝑛𝑞𝑝

∞ in probability 𝑃0

.∎

Lemma(4):
Assuming that the true model is the mixed one-way repeated measurements model (the model under 𝐻1 in 5). Let 𝑌𝑛𝑞𝑝
= (𝑦1 , … , 𝑦𝑛𝑞𝑝 ) where 𝑦𝑖𝑗𝑘 ′𝑠 are independent normal random variables with mean μ + τj + δi(j) + γk + 𝑇

(τγ)jk and variance 𝜎𝜖2 = 1 and let 𝑤𝑛𝑞𝑝 = 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 ).
Then, lim𝑛𝑞𝑝→∞

1 𝑌𝑇
𝑄 𝑌 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

= 1, in probability 𝑃1∞ .

Proof:
From the moment formula of the quadratic form of normal random variables, the expectation and variance of 𝑇

quadratic form 𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 are given by: 𝑇 𝑇 𝐸

(𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) = 𝑡𝑟 (𝑄𝑛𝑞𝑝 ) + 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 ) 𝑛𝑞𝑝
−(1+𝑞+𝑝+𝑞𝑝)

= ∑𝑖=1 𝑛𝑞𝑝𝜎𝛿

2 𝑇

(1+𝑛𝑞𝑝𝜎𝛿2 )

+ 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 )

= 𝑆𝑛𝑞𝑝,1 + 𝑤𝑛𝑞𝑝 , and 𝑇 𝑇 𝑣𝑎𝑟

(𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) = 2𝑡𝑟 (𝑄𝑛𝑞𝑝 2 ) + 4𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 2 𝐸(𝑌𝑛𝑞𝑝 ).
Note 𝑇 𝑤𝑛𝑞𝑝

= 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 ) 𝑇 𝑇

= (𝜇, 𝜏 𝑇 , 𝛾 𝑇 , (𝜏𝛾)𝑇 , 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
)𝑀𝑛𝑞𝑝

0
0𝑞×1

01×𝑞
0𝑞×𝑞

0𝑝×1

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1
[

0𝑛𝑞×𝑞

01×𝑝
0𝑞×𝑝
0𝑝×𝑝
0𝑞𝑝×𝑝

01×𝑞𝑝
0𝑞×𝑞𝑝
0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

01×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

(

2
(𝑛𝑞𝑝) 𝜎2𝛿
2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿 )𝜎2𝜖

0
0𝑞×1

=

0 𝑇

(𝜇, 𝜏 𝑇 , 𝛾 𝑇 , (𝜏𝛾)𝑇 , 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
) 𝑝×1 𝑇 𝑇 𝑇

(𝜇, 𝜏 , 𝛾 , (𝜏𝛾)

1 𝑀 𝑛𝑞𝑝

2 𝑛𝑞𝑝 𝑇

]
01×𝑝

0𝑞×𝑞
0𝑝×𝑞

0𝑞×𝑝

0𝑞𝑝×𝑞

[0𝑛𝑞×1

0𝑛𝑞×𝑞

01×𝑞𝑝

0𝑝×𝑝

0𝑝×𝑞𝑝

0𝑞𝑝×𝑝

0𝑞𝑝×𝑞𝑝

0𝑛𝑞×𝑝

(𝑛𝑞𝑝)2 𝜎𝛿2
2
(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

01×𝑛𝑞

0𝑞×𝑞𝑝

0𝑛𝑞×𝑞𝑝 𝑇 𝑇

, 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
) 𝑇

= 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
( 𝑇

)𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

01×𝑞

0𝑞𝑝×1 𝑇 𝑀𝑇𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 (𝜇, 𝜏𝑇 , 𝛾𝑇 , (𝜏𝛾) , 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) )

0𝑞×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

(

(𝑛𝑞𝑝)2 𝜎𝛿2

(𝜎𝜖2 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2

) 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)
) 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) = ∑𝑖=1

]

(𝑛𝑞𝑝)2 𝜎𝛿2 𝛿𝑖2
1+𝑛𝑞𝑝𝜎𝛿2

,

since 𝑠𝑢𝑝𝑖 𝛿𝑖 < ∞ , we have

2 𝑛𝑞𝑝
𝑛𝑞𝑝𝜎𝛿
1+𝑛𝑞𝑝𝜎𝛿2
2
consider 𝛿𝑚
>0, 𝑤𝑛𝑞𝑝

≤ 𝑠𝑢𝑝𝑖 𝛿𝑖2 𝑛𝑞𝑝 ∑𝑖=1

= 𝑂(𝑛𝑞𝑝 𝑆𝑛𝑞𝑝,1 ), then 𝑤𝑛𝑞𝑝 = 𝑂(𝑛𝑞𝑝2 ).

Now we can

where 𝑚 ≥ 1 is a fixed positive integer and a sufficiently large 𝑛𝑞𝑝 with 𝑛𝑞𝑝𝜎𝛿

2

1+𝑛𝑞𝑝𝜎𝛿2

1
2

. Then for such 𝑛𝑞𝑝 and 𝑚. 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝑤𝑛𝑞𝑝
= ∑𝑖=1

= 𝑛𝑞𝑝 𝑛𝑞𝑝𝜎𝛿

2

1+𝑛𝑞𝑝𝜎𝛿2 𝑛𝑞𝑝

2 𝜎𝛿2 𝛿𝑖2

2 𝛿𝑚

≥ 𝑛𝑞𝑝

1+𝑛𝑞𝑝𝜎𝛿2
2 𝛿𝑚

2

≥ 𝑛𝑞𝑝

2 𝜎𝛿2

1+𝑛𝑞𝑝𝜎𝛿2

2 𝛿𝑚

.

Similarly to the previous calculation, we have 𝑇 𝑇

2 𝑇 𝐸

(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸
(𝑌𝑛𝑞𝑝 ) = (𝜇, 𝜏 𝑇 , 𝛾 𝑇 , (𝜏𝛾)𝑇 , 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
)𝑀𝑛𝑞𝑝

153

1 𝑀

(𝑛𝑞𝑝)3 𝑛𝑞𝑝

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014
0
0𝑞×1

01×𝑞
0𝑞×𝑞

0𝑝×1

0𝑝×𝑞

0𝑞𝑝×1

0𝑞𝑝×𝑞

0𝑛𝑞×1
[

0𝑛𝑞×𝑞

01×𝑝
0𝑞×𝑝
0𝑝×𝑝
0𝑞𝑝×𝑝

01×𝑞𝑝
0𝑞×𝑞𝑝
0𝑝×𝑞𝑝
0𝑞𝑝×𝑞𝑝

0𝑛𝑞×𝑝

0𝑛𝑞×𝑞𝑝

=
=

www.iiste.org

01×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞×𝑛𝑞
0𝑞𝑝×𝑛𝑞

( 𝑇

) 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

2

(𝜎𝜖 +𝑛𝑞𝑝𝜎𝛿 )𝜎2𝜖

]

(𝑛𝑞𝑝)2 𝜎𝛿2

1 𝛿𝑇

( 𝑛𝑞𝑝
𝑛𝑞𝑝−((1+𝑞+𝑝+𝑞𝑝) (𝜎𝜖2 +𝑛𝑞𝑝𝜎𝛿2 )𝜎𝜖2
1

∑𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑖
=1 𝑛𝑞𝑝 𝑇

2
Since 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸
(𝑌𝑛𝑞𝑝 ) ≤ ∑ 𝑇 𝑇

2

2

(𝑛𝑞𝑝) 𝜎2𝛿
2 𝑇 𝑀𝑇𝑛𝑞𝑝

𝑀𝑛𝑞𝑝 (𝜇, 𝜏𝑇 , 𝛾𝑇 , (𝜏𝛾) , 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) )

(𝑛𝑞𝑝)4 𝜎𝛿4 𝛿𝑖2

(1+𝑛𝑞𝑝𝜎𝛿2 )2

3 4 2 𝑛𝑞𝑝
(𝑛𝑞𝑝) 𝜎𝛿 𝛿𝑖
2 𝑖
=1
(1+𝑛𝑞𝑝𝜎𝛿2 )

)2 𝐼𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝛿𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)

.
2 4 𝑛𝑞𝑝
(𝑛𝑞𝑝) 𝜎𝛿

≤ 𝑛𝑞𝑝 𝑠𝑢𝑝𝑖 𝛿𝑖2 ∑𝑖=1

2

(1+𝑛𝑞𝑝𝜎𝛿2 )

= 𝑂( 𝑛𝑞𝑝𝑆𝑛𝑞𝑝,2 ) (as in case

of 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 )).
Now for previous 𝑛𝑞𝑝 and 𝑚 we can find 𝑇 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)
2 𝐸
(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸
(𝑌𝑛𝑞𝑝 ) = ∑𝑖=1

(𝑛𝑞𝑝)3 𝜎𝛿4 𝛿𝑖2

2
(1+𝑛𝑞𝑝𝜎𝛿2 )

2
≥ 𝑛𝑞𝑝 𝛿𝑚
( 𝑛𝑞𝑝𝜎𝛿

2

2


2)

1+𝑛𝑞𝑝𝜎𝛿

2 𝛿𝑚 𝑛𝑞𝑝

.

4

By combining value of 𝑆𝑛𝑞𝑝,1 and the previous result, get 𝑇 𝑣𝑎𝑟

(𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) ≤ 𝑂(𝑛𝑞𝑝 𝑆𝑛𝑞𝑝,1 ). 𝑇 𝑇

Furthermore, note that 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 𝐸(𝑌𝑛𝑞𝑝 ) ≥ 𝐸(𝑌𝑛𝑞𝑝 ) 𝑄𝑛𝑞𝑝 2 𝐸(𝑌𝑛𝑞𝑝 ).
By the Chebshev inequality[13] get 𝑝𝑟
{|

1 𝑌𝑇
𝑄 𝑌 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

Since lim𝑛𝑞𝑝→∞
∴ lim𝑛𝑞𝑝→∞

− 𝑇

𝑄 𝐸
(𝑌𝑛𝑞𝑝 𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) 𝑇

𝑄 𝑣𝑎𝑟
(𝑌𝑛𝑞𝑝 𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) 𝑤𝑛𝑞𝑝

2 𝜀2 𝑤𝑛𝑞𝑝 𝑇

𝑄 𝐸
(𝑌𝑛𝑞𝑝 𝑛𝑞𝑝
𝑌𝑛𝑞𝑝 ) 𝑤𝑛𝑞𝑝

1 𝑌𝑇
𝑄 𝑌 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

|> 𝜀}≤ 𝑆𝑛𝑞𝑝
,1

= lim𝑛𝑞𝑝→∞ ( 𝑤𝑛𝑞𝑝

+

= 1 , in probability 𝑤𝑛𝑞𝑝 𝑤𝑛𝑞𝑝 𝑛𝑞𝑝

→∞

= 𝑂(𝑛𝑞𝑝−2 ) →

0, where 𝜀 > 0 .

)=1

Lemma(5):
Suppose we have the mixed one- way repeated measurements model, then
lim𝑛𝑞𝑝→∞

1 𝑤𝑛𝑞𝑝

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )

=0

Proof: 𝑛𝑞𝑝

2
2
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) ≤ ∏𝑖=1 (𝜎𝜖2 + 𝜎2𝜇 𝑥2𝑖 + ∑𝑞𝑘=1 𝜎2𝜏 𝑥2𝑖𝑘 + ∑𝑝𝑘=1 𝜎2𝛾 𝑥2𝑖𝑘 + ∑𝑞𝑝 𝑘
=1 𝜎(𝜏𝛾) 𝑥𝑖𝑘 +

Since

2 2
∑𝑛𝑞 𝑘
=1 𝜎𝛿 𝑥𝑖𝑘 ) (from the property of the determinant of the positive definite matrix) 𝑛𝑞𝑝

= (𝑛𝑞𝑝)𝑛𝑞𝑝 ∏𝑖=1 ( 𝜎𝜖

2 𝑛𝑞𝑝 𝑞

+ 𝑝 𝑞𝑝 𝑛𝑞

2
2
2 2
2 𝑥 2 +∑
2 2
2 2
(𝜎𝜇 𝑖 𝑘

=1 𝜎𝜏 𝑥𝑖𝑘 +∑𝑘=1 𝜎𝛾 𝑥𝑖𝑘 +∑𝑘=1 𝜎(𝜏𝛾) 𝑥𝑖𝑘 +∑𝑘=1 𝜎𝛿 𝑥𝑖𝑘 𝑛𝑞𝑝

).

And we have
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) = (𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝 𝜎𝛿

2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) (from lemma
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )
→ log

+ 𝜎𝜇2 )( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝜏2 )𝑞 ( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛾2 )𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
)𝑞𝑝 ( 𝜎𝜖

2 𝑛𝑞𝑝

+

1).

= log det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) − log det(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 +

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟∑1,𝑟 𝑍𝑟𝑇 ) 𝑍𝑟

∑1,𝑟 𝑍𝑟𝑇 ) 𝑛𝑞𝑝

≤ {(𝑛𝑞𝑝)log(𝑛𝑞𝑝) + ∑𝑖=1 log ( 𝜎𝜖

2 𝑛𝑞𝑝

log ( 𝜎𝜖

2 𝑛𝑞𝑝 𝑞𝑝

)log (

+ 𝜎𝜇2 ) + 𝑞log ( 𝜎𝜖
2 𝑛𝑞𝑝 𝜎𝜖

2 𝑛𝑞𝑝 𝑞

+ 𝑝 𝑞𝑝 𝑛𝑞

2
2
2 2
2 𝑥 2 +∑
2 2
2 2
(𝜎𝜇 𝑖 𝑘

=1 𝜎𝜏 𝑥𝑖𝑘 +∑𝑘=1 𝜎𝛾 𝑥𝑖𝑘 +∑𝑘=1 𝜎(𝜏𝛾) 𝑥𝑖𝑘 +∑𝑘=1 𝜎𝛿 𝑥𝑖𝑘 𝑛𝑞𝑝

+ 𝜎𝜏2 ) + 𝑝log ( 𝜎𝜖

2 𝑛𝑞𝑝

+ 𝜎𝛾2 ) + 𝑞𝑝log ( 𝜎𝜖

2 𝑛𝑞𝑝

2
+ 𝜎(𝜏𝛾)
) + 𝑛𝑞𝑝 − (1 + 𝑞 + 𝑝 +

+ 𝜎𝛿2 )} = 𝑂(1) − 𝑂(1) = 𝑂(1).

Then
1 𝑤𝑛𝑞𝑝

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 ) 𝑛𝑞𝑝→∞
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )

Theorem(2):
154

)} − {𝑛𝑞𝑝log(𝑛𝑞𝑝) +

0

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

Suppose that the true model is the mixed one- way repeated measurements model (the model under 𝐻1 in 5), 𝑛𝑞𝑝

and let 𝑃1
denote the true distribution of the data, with probability density function 𝑝
1 (yijk |μτj , δi(j) , γk , (τγ)jk ) = 𝛷{(yijk − (μ + τj + δi(j) + γk + (τγ)jk ))/𝜎𝜖 }, where 𝛷(∙) is the standard normal
density. Then the Bayes factor is consistent under the alternative hypothesis 𝐻1 i.e 𝑛𝑞𝑝

lim𝑛𝑞𝑝→∞ 𝐵01 = 0 in probability 𝑃1 .

Proof:
1

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

log 𝐵01 = log

−1

1 𝑇

− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) )𝑌𝑛𝑞𝑝
2

=
1

log

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 ) 𝑇 𝑋

∑0 𝑋 +

−1 𝑍𝑟
∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝

1

+

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )

2

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

1 𝑌

𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝
2 𝑛𝑞𝑝 𝑇

+ 𝑋∑0 𝑋 +

1 𝑇

− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 +

2 𝑇
−1 𝑍𝑟
∑1,𝑟 𝑍𝑟 ) 𝑇

−1

−(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 ) )𝑌𝑛𝑞𝑝
1 𝑇
1 𝑇
−1
= log𝐵̃01 − 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝 − 𝑌𝑛𝑞𝑝
2
2
−1
−1
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) )𝑌𝑛𝑞𝑝 , where
log𝐵̃01 =
1
2

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

−1

1 𝑇

− 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝 .
2

From lemma (1), we have
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟∑1,𝑟 𝑍𝑟𝑇 )

2
2
2
2 𝜎
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 ( 𝜎𝜖 +𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝛿 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝜎

2 𝜎
2 𝜎
2 𝜎
2
2 𝜖 𝜖 𝜖 𝜖

2
2
2
1+𝑞+𝑝+𝑞𝑝 𝑞 𝑝 𝑛𝑞𝑝

(
+𝜎𝜇 )(
+𝜎𝜏 ) (
+𝜎𝛾 ) (
+𝜎(𝜏𝛾) )𝑞𝑝 (𝜎𝜖2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝜎
2 𝜖
+𝜎 2 ) 𝑛𝑞𝑝
−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝𝑛𝑞𝑝
−(1+𝑞+𝑝+𝑞𝑝) (𝑛𝑞𝑝 𝛿 𝜎𝜖

2 +𝑛𝑞𝑝𝜎𝛿2 𝑛𝑞𝑝
−(1+𝑞+𝑝+𝑞𝑝) 𝜎𝜖
2 𝜎𝜖
2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

=

=

=(

=(1 + 𝑛𝑞𝑝𝜎𝛿

2 𝜎𝜖
2

)

⟹ log (1 + 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝) 𝑛𝑞𝑝𝜎𝛿

2 𝜎𝜖
2

)

)

= 𝑂(𝑛𝑞𝑝) 𝑛𝑞𝑝

−(1+𝑞+𝑝+𝑞𝑝)

= 𝑂(𝑛𝑞𝑝).

And from lemma (4) 𝑤𝑛𝑞𝑝 = 𝑂(𝑛𝑞𝑝2 ). Then
1 𝑤𝑛𝑞𝑝

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 ) 𝑛𝑞𝑝

→∞

= 𝑂(𝑛𝑞𝑝−1 ) →

0.

Then,
1 𝑤𝑛𝑞𝑝

log𝐵̃01 =

1

1

(

2 𝑤𝑛𝑞𝑝

log

det(𝑉1,𝑛𝑞𝑝 )

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

1 𝑌
𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝

+ 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 +

−1 𝑍𝑟

∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝 ).
−1

Since (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 ) −1 − (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) 𝑌
𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝
1
(4), lim𝑛𝑞𝑝→∞ 𝑤

+ 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝 = 𝑛𝑞𝑝

1 𝑤𝑛𝑞𝑝 𝑛𝑞𝑝

→∞

log𝐵̃01 →

1

(

= 𝑛𝑞𝑝

→∞ −1 𝑤𝑛𝑞𝑝

1

log

2 𝑤𝑛𝑞𝑝 𝑛𝑞𝑝
→∞
1 𝑍𝑟

∑1,𝑟 𝑍𝑟𝑇 ) )𝑌𝑛𝑞𝑝 ) →
1

1 𝑌𝑇
𝑄 𝑌 . 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

from lemma 𝑇 𝑌𝑛𝑞𝑝 𝑄𝑛𝑞𝑝

𝑌𝑛𝑞𝑝 = 1 , then

=

−1

= 𝑄𝑛𝑞𝑝 , then
−1

1

log𝐵̃01 →

2

2

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 )

(0 − 1) =

, in probability

1 𝑌
𝑇 ((𝜎𝜖2 𝐼𝑛𝑞𝑝 𝑤𝑛𝑞𝑝
𝑛𝑞𝑝

+ 𝑋∑0 𝑋 𝑇 ) −1 −(𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 +

−1
2 𝑛𝑞𝑝 𝑃

1 .

Now from lemma (5),
1

1 𝑤𝑛𝑞𝑝

2

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) 𝑛𝑞𝑝

→∞

1

−1 𝑇

0, and − 𝑌𝑛𝑞𝑝
((𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 ) −(𝜎𝜖2 𝐼𝑛𝑞𝑝 +
2

−1 𝑋

∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 ) )𝑌𝑛𝑞𝑝 is nonpositive random variable since (𝜎𝜖2 𝐼𝑛𝑞𝑝 + 𝑋∑0 𝑋 𝑇 + 𝑍𝑟 ∑1,𝑟 𝑍𝑟𝑇 )
(𝜎𝜖2 𝐼𝑛𝑞𝑝 𝑇 𝑇

−1

+ 𝑋∑0 𝑋 + 𝑍∑1 𝑍 )

is a nonnegative definite matrix.

155

−1

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

Therefore, by combining the above results, there exist 𝐶2 > 0 such that
1

lim𝑛𝑞𝑝→∞ sup 𝑤𝑛𝑞𝑝 𝑛𝑞𝑝

→∞

Then , 𝐵01 →

−𝐶2

log 𝐵01 ≤

2

, with probability tending to 1. 𝑛𝑞𝑝

0, in probability 𝑃1

.∎

4.Conclusions
1- The posterior distribution 𝜋1 (𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 ) is the multivariate normal with 𝐸
(𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 , 𝜎𝜖2 ) = ∑2 (∑2 + 𝜎𝜖2 𝐼𝑁 )−1 𝑌𝑛𝑞𝑝 , where N=1+q+p+qp+nq and
2
−1 −1 𝑉𝑎𝑟
(𝜉𝑛𝑞𝑝 |𝑌𝑛𝑞𝑝 , 𝜎𝜖2 ) = (∑−1
= ∑2 (∑2 + 𝜎𝜖2 𝐼𝑁 )−1 𝜎𝜖2 𝐼𝑁 , where
2 + (𝜎𝜖 𝐼𝑁 ) ) 𝜉𝑖𝑗𝑘
= μ + τj + δi(j) + γk + (τγ)jk + eijk ,for, 𝑖 = 1,2, … , 𝑛, 𝑗 = 1, … , 𝑞 𝑎𝑛𝑑 𝑘 = 1, … , 𝑝, and

∑2 = 𝑋∑0 𝑋 𝑇 + 𝑍∑1 𝑍 𝑇 = 𝑋
[ 𝜎𝛿
211

0

⋯ 𝜎𝜇

2

01×𝑞

01×𝑝

0𝑞×1 𝜎𝜏

2 𝐼𝑞×𝑞

0𝑞×𝑝

0𝑝×1
0𝑞𝑝×1

0𝑝×𝑞
0𝑞𝑝×𝑞 𝜎𝛾

2 𝐼𝑝×𝑝

01×𝑞𝑝
0𝑞×𝑞𝑝
0𝑝×𝑞𝑝 𝑋𝑇

+ 𝑍

2 𝜎
(𝜏𝛾) 𝐼𝑞𝑝
×𝑞𝑝 ]

0𝑞𝑝×𝑝

0 𝜎𝛿

212

0

⋯ 0 𝑍𝑇





⋯ 𝜎𝛿2𝑛𝑞 ]
0
[ 0
2-The Bayes factor for testing problem 𝐻0 : yijk = μ + τj + γk + (τγ)jk + eijk versus 𝐻
1 : yijk = μ + τj + δi(j) + γk + (τγ)jk + eijk is given by: 𝑞 𝐵

01 = 𝑝

2
2
2
2
2 )( 𝜎𝜖 +𝜎 2 ) ( 𝜎𝜖 +𝜎 2 ) ( 𝜎𝜖 +𝜎 2 )
√(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑞𝑝 𝑛𝑞𝑝

2 𝜎

( 𝜖 +𝜎𝛿2 )𝑛𝑞 𝑛𝑞𝑝

2
2
2
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 (𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾 𝜖

(𝜏𝛾) 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝜎𝜖

2

(𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝

−𝑀𝑛𝑞𝑝 [ 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐺𝑛𝑞𝑝 ]

−1

exp{ −

1 𝑌𝑇

2𝑛𝑞𝑝2 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

)𝑌𝑛𝑞𝑝 }.

3-The approximation of 𝐵01 with design matrix of the random effects (𝑍) to the first 𝑛𝑞𝑝 terms as following
2 𝐵

̃01 =

2

2

2

2

2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 ( 𝜎𝜖 +𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√(𝑛𝑞𝑝)𝑛𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾

(𝜏𝛾) 𝛿 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑛𝑞𝑝

2
2
2
2
2 )( 𝜎𝜖 +𝜎 2 )𝑞 ( 𝜎𝜖 +𝜎 2 )𝑝 ( 𝜎𝜖 +𝜎 2 )𝑞𝑝 (𝜎 2 )𝑛𝑞𝑝−(1+𝑞+𝑝+𝑞𝑝)
√𝑛𝑞𝑝1+𝑞+𝑝+𝑞𝑝 ( 𝜎𝜖 +𝜎𝜇 𝜏 𝛾 𝜖

(𝜏𝛾) 𝑛𝑞𝑝

(𝑀𝑛𝑞𝑝 [ 𝑛𝑞𝑝

−1 𝜎𝜖

2 𝑛𝑞𝑝 𝐼𝑛𝑞𝑝

+ 𝐶𝑛𝑞𝑝 ] 𝑛𝑞𝑝

exp{−

1 𝑌𝑇

2𝑛𝑞𝑝2 𝑛𝑞𝑝 𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

−𝑀𝑛𝑞𝑝 [ 𝜎𝜖

2 𝑛𝑞𝑝

−1 𝐼𝑛𝑞𝑝

+ 𝐷𝑛𝑞𝑝 ] 𝑇 𝑀𝑛𝑞𝑝

)𝑌𝑛𝑞𝑝 }.

4- If we have the one- way repeated measurements model, then
1 𝑆𝑛𝑞𝑝
,1 𝑡𝑟

(

1 𝑛𝑞𝑝

2 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑀𝑛𝑞𝑝

[

−1

+ 𝐷𝑛𝑞𝑝 ] 𝜎𝜖

2 𝐼 𝑛𝑞𝑝

𝑛𝑞𝑝 𝑇 𝑀𝑛𝑞𝑝

− [𝑋 𝑍] [

−1

+ 𝐺𝑛𝑞𝑝 ] 𝑛𝑞𝑝

→∞

[𝑋 𝑍]𝑇 ) →

0

5-If the true model is the one-way repeated measurement model with fixed effects only then the Bayes factor is 𝑛𝑞𝑝

consistent under the null hypothesis 𝐻0 . This implies that, lim𝑛𝑞𝑝→∞ 𝐵01 = ∞ in probability 𝑃0 .
6- If we have the mixed one- way repeated measurements model, then
lim𝑛𝑞𝑝→∞

1 𝑤𝑛𝑞𝑝

log

det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍∑1 𝑍 𝑇 )
det(𝜎𝜖2 𝐼𝑛𝑞𝑝 +𝑋∑0 𝑋 𝑇 +𝑍𝑟 ∑1,𝑟𝑍𝑟𝑇 )

= 0.

7- If the true model is the mixed one- way repeated measurements model, then the Bayes factor is consistent 𝑛𝑞𝑝

under the alternative hypothesis 𝐻1 .This implies that, lim𝑛𝑞𝑝→∞ 𝐵01 = 0 in probability 𝑃1 .

5.References
[1] Aerts, M. , Claeskens, G. and Hart, J.D., (2004). Bayesian –motivated tests of function fit and their
asymptotic frequentist properties, Ann. Statist. 32 (6), 2580-2615.
[2] Al-Mouel, A.H.S.,(2004) . Multivariate Repeated Measures Models and Comparison of Estimators, h. D.
Thesis, East China Normal University, China.
[3] Al-Mouel , A.H.S., and Wang, J.L,(2004). One –Way Multivariate Repeated Measurements
analysis of
Variance Model, Applied Mathematics a Journal of Chinese Universities, 19,4,pp435448.

156

Mathematical Theory and Modeling
ISSN 2224-5804 (Paper) ISSN 2225-0522 (Online)
Vol.4, No.11, 2014

www.iiste.org

[4] Arnold, Steven F., (1981). The theory of linear models and multivariate analysis. John Wiley and
Sons, New York. Chichester. Brisbane. Toronto .
[5] Berger, J. O., (1985). Statistical Decision Theory and Bayesian Analysis. Springer, New York.
[6] Choi, T., Lee, J. and Roy, A., (2008). A note on the Bayes factor in a semiparametric regression
model, Journal of Multivariate Analysis. 1316-1327.
[7] Dass, S.C. and Lee, J., (2004). A note on the consistency of Bayes factors for testing point null
versus
non-parametric alternatives, J. Statist. plan. Inference 119 (1) 143-152.
[8] Ghosh, J.K. , Delampady, M. and Samanta, T., (2006). Introduction to Bayesian Analysis: Theory and
Methods, Springer, New York.
[9] Gosal, G., Lember, J. and Van der Vaart, A., (2008). Nonparametric Bayesian model selection and
averaging, Electron. J. Stat. 2 63-89.
[10] Graybill, F.A., (1976). Theory and Application of the linear Model, Duxbury press, North Scituate, Mass,
xiv+704.
[11] Mohaisen, A. J. and Abdulhussain , A. M., (2013). Large Sample Property of The Bayes Factor in a Spline
Semiparametric Regression Model , Mathematical Theory and Modeling www.iiste.org ISSN
(Paper)2224-5804 ISSN (Online)2225-0522 Vol 3, No. 11, 2013.
[12] Neter, J. and Wasserman, W. , (1974). Applied linear statistical models, regression analysis of
variance
and experimental designs, Richard. D. IRWIN, INC .
[13] Papoulis,A.,(1997). Markoff Sequences in Probability, Random Variables, and Stochastic
Processes,
2nd ed. New York . McGraw-Hill, PP. 528-535, 1984.
[14] Tong, Y.L., (1989). The Multivariate Normal Distribution, Springer series in statistics, Springer.
[15] Verdinelli, I. and Wasserman, L., (1998). Bayesian goodness –of-fit testing using infinitedimensional
exponential families, Ann. Stat.26 (4) 1215-1241.
[16] Vinish, R. Mc , Rousseau, J. and Mengerson, K., (2008). Bayesian goodness –of-fit testing with
mixtures of triangular distributions, Scand. J, statist. in press (doi:10.1111/j.14679469.2008.00620.x).
[17] Vonesh, E.F. and Chinchilli, V.M., (1997). Linear and Nonlinear Models for the Analysis of
Repeated
Measurements, Marcel Dakker, Inc., New York.

157

Business, Economics, Finance and Management Journals
European Journal of Business and Management
Research Journal of Finance and Accounting
Journal of Economics and Sustainable Development
Information and Knowledge Management
Journal of Developing Country Studies
Industrial Engineering Letters

PAPER SUBMISSION EMAIL
EJBM@iiste.org
RJFA@iiste.org
JESD@iiste.org
IKM@iiste.org
DCS@iiste.org
IEL@iiste.org

Physical Sciences, Mathematics and Chemistry Journals
Journal of Natural Sciences Research
Journal of Chemistry and Materials Research
Journal of Mathematical Theory and Modeling
Advances in Physics Theories and Applications
Chemical and Process Engineering Research

PAPER SUBMISSION EMAIL
JNSR@iiste.org
CMR@iiste.org
MTM@iiste.org
APTA@iiste.org
CPER@iiste.org

Engineering, Technology and Systems Journals
Computer Engineering and Intelligent Systems
Innovative Systems Design and Engineering
Journal of Energy Technologies and Policy
Information and Knowledge Management
Journal of Control Theory and Informatics
Journal of Information Engineering and Applications
Industrial Engineering Letters
Journal of Network and Complex Systems

PAPER SUBMISSION EMAIL
CEIS@iiste.org
ISDE@iiste.org
JETP@iiste.org
IKM@iiste.org
CTI@iiste.org
JIEA@iiste.org
IEL@iiste.org
NCS@iiste.org

Environment, Civil, Materials Sciences Journals
Journal of Environment and Earth Science
Journal of Civil and Environmental Research
Journal of Natural Sciences Research

PAPER SUBMISSION EMAIL
JEES@iiste.org
CER@iiste.org
JNSR@iiste.org

Life Science, Food and Medical Sciences
Advances in Life Science and Technology
Journal of Natural Sciences Research
Journal of Biology, Agriculture and Healthcare
Journal of Food Science and Quality Management
Journal of Chemistry and Materials Research

PAPER SUBMISSION EMAIL
ALST@iiste.org
JNSR@iiste.org
JBAH@iiste.org
FSQM@iiste.org
CMR@iiste.org

Education, and other Social Sciences
Journal of Education and Practice
Journal of Law, Policy and Globalization
Journal of New Media and Mass Communication
Journal of Energy Technologies and Policy
Historical Research Letter
Public Policy and Administration Research
International Affairs and Global Strategy
Research on Humanities and Social Sciences
Journal of Developing Country Studies
Journal of Arts and Design Studies

PAPER SUBMISSION EMAIL
JEP@iiste.org
JLPG@iiste.org
NMMC@iiste.org
JETP@iiste.org
HRL@iiste.org
PPAR@iiste.org
IAGS@iiste.org
RHSS@iiste.org
DCS@iiste.org
ADS@iiste.org

The IISTE is a pioneer in the Open-Access hosting service and academic event management.
The aim of the firm is Accelerating Global Knowledge Sharing.
More information about the firm can be found on the homepage:
http://www.iiste.org
CALL FOR JOURNAL PAPERS
There are more than 30 peer-reviewed academic journals hosted under the hosting platform.
Prospective authors of journals can find the submission instruction on the following
page: http://www.iiste.org/journals/ All the journals articles are available online to the
readers all over the world without financial, legal, or technical barriers other than those
inseparable from gaining access to the internet itself. Paper version of the journals is also
available upon request of readers and authors.

MORE RESOURCES
Book publication information: http://www.iiste.org/book/

IISTE Knowledge Sharing Partners
EBSCO, Index Copernicus, Ulrich's Periodicals Directory, JournalTOCS, PKP Open
Archives Harvester, Bielefeld Academic Search Engine, Elektronische Zeitschriftenbibliothek
EZB, Open J-Gate, OCLC WorldCat, Universe Digtial Library , NewJour, Google Scholar