• Nenhum resultado encontrado

Hypothesis testing in variance components with constraints

N/A
N/A
Protected

Academic year: 2023

Share "Hypothesis testing in variance components with constraints"

Copied!
83
0
0

Texto

(1)

Hypothesis testing in variance components with constraints

M. Przystalski1 M. Fonseca2

1The Research Center for Cultivar Testing, S lupia Wielka, Poland

2Centro de Mathem´atica e Aplica¸oes, Faculdade de Ciˆencias e Tecnologia, Universidade Nova de Lisboa, Lisboba, Portugal

LinStat 2014

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 1 / 32

(2)

Introduction

For testing hypotheses related to variance components one uses e.g.

permutation test (e.g. Fitzmaurice at al., 2007), or generalized p-values (see e.g. Volaufova, 2009).

For inference on variance component the usual chi-square distribution of likelihood ratio statistic under the null is used.

E.g. Self and Liang (1987) had shown that this not necessarily hold. For LMM with one variance component, Crainiceanu and Ruppert (2004) had shown that the distribution of the LR statistic is mixture of two chi-squares.

This is due to the fact that the values of the variance component under the null lie on the boundry of the parameter space.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 2 / 32

(3)

Introduction

For testing hypotheses related to variance components one uses e.g.

permutation test (e.g. Fitzmaurice at al., 2007), or generalized p-values (see e.g. Volaufova, 2009).

For inference on variance component the usual chi-square distribution of likelihood ratio statistic under the null is used.

E.g. Self and Liang (1987) had shown that this not necessarily hold. For LMM with one variance component, Crainiceanu and Ruppert (2004) had shown that the distribution of the LR statistic is mixture of two chi-squares.

This is due to the fact that the values of the variance component under the null lie on the boundry of the parameter space.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 2 / 32

(4)

Introduction

For testing hypotheses related to variance components one uses e.g.

permutation test (e.g. Fitzmaurice at al., 2007), or generalized p-values (see e.g. Volaufova, 2009).

For inference on variance component the usual chi-square distribution of likelihood ratio statistic under the null is used.

E.g. Self and Liang (1987) had shown that this not necessarily hold.

For LMM with one variance component, Crainiceanu and Ruppert (2004) had shown that the distribution of the LR statistic is mixture of two chi-squares.

This is due to the fact that the values of the variance component under the null lie on the boundry of the parameter space.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 2 / 32

(5)

Introduction

For testing hypotheses related to variance components one uses e.g.

permutation test (e.g. Fitzmaurice at al., 2007), or generalized p-values (see e.g. Volaufova, 2009).

For inference on variance component the usual chi-square distribution of likelihood ratio statistic under the null is used.

E.g. Self and Liang (1987) had shown that this not necessarily hold.

For LMM with one variance component, Crainiceanu and Ruppert (2004) had shown that the distribution of the LR statistic is mixture of two chi-squares.

This is due to the fact that the values of the variance component under the null lie on the boundry of the parameter space.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 2 / 32

(6)

Introduction

For testing hypotheses related to variance components one uses e.g.

permutation test (e.g. Fitzmaurice at al., 2007), or generalized p-values (see e.g. Volaufova, 2009).

For inference on variance component the usual chi-square distribution of likelihood ratio statistic under the null is used.

E.g. Self and Liang (1987) had shown that this not necessarily hold.

For LMM with one variance component, Crainiceanu and Ruppert (2004) had shown that the distribution of the LR statistic is mixture of two chi-squares.

This is due to the fact that the values of the variance component under the null lie on the boundry of the parameter space.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 2 / 32

(7)

Introduction

Aims

1 Construct statiscal tests for variance components with posistivity constraints.

2 Investigate the perfomance of the proposed procedures.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 3 / 32

(8)

Introduction

Aims

1 Construct statiscal tests for variance components with posistivity constraints.

2 Investigate the perfomance of the proposed procedures.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 3 / 32

(9)

One-way model

Statisitcal model

y =µ1+Zα+e, where

y - vector of observed values, µ- general mean

1N - vector of one’s

α - vector of random effects, Z=Ia⊗1n is design matrix, e - vector of errors

We assume that α ande are indendent and that α∼N 0, σa2Ia and e ∼N 0, σ2eIN

,respectively.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 4 / 32

(10)

One-way model

Statistical model

The problem of interest is:

H02a = 0 vs. H12a >0.

Alternatively, the problem of interest can be written as: H0012 vs. H101 > γ2,

whereγ1 =nσ2ae2 andγ2e2 (see Khuri et al., 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 5 / 32

(11)

One-way model

Statistical model

The problem of interest is:

H02a = 0 vs. H12a >0.

Alternatively, the problem of interest can be written as:

H0012 vs. H101 > γ2,

whereγ1=nσ2a2e andγ2e2 (see Khuri et al., 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 5 / 32

(12)

One-way model

Statistical model

The problem of interest is:

H02a = 0 vs. H12a >0.

Alternatively, the problem of interest can be written as:

H0012 vs. H101 > γ2, whereγ1=nσ2a2e andγ2e2 (see Khuri et al., 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 5 / 32

(13)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n .

E(bγ1) =γ1 andV(bγ1)≈ a−121;

E(bγ2) =γ2 and; V(γb2)≈ N−a22

. By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have

b =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(14)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n . E(bγ1) =γ1 andV(bγ1)≈ a−121;

E(bγ2) =γ2 and; V(γb2)≈ N−a22

.

By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have

b =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(15)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n . E(bγ1) =γ1 andV(bγ1)≈ a−121;E(bγ2) =γ2 and; V(γb2)≈ N−a22 .

By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have

b =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(16)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n . E(bγ1) =γ1 andV(bγ1)≈ a−121;E(bγ2) =γ2 and; V(γb2)≈ N−a22 . By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have

b =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(17)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n . E(bγ1) =γ1 andV(bγ1)≈ a−121;E(bγ2) =γ2 and; V(γb2)≈ N−a22 . By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have V

γb =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(18)

One-way model

Statistical model

The maximum likelihood estimators (MLE)

bµ= 1

N10Ny, γb1 = 1

a−1y0P1y bγ2= 1

N−ay0P2y, where

P1 = Ia⊗Jn n − JN

N P2=Ia⊗In−Ia⊗Jn n . E(bγ1) =γ1 andV(bγ1)≈ a−121;E(bγ2) =γ2 and; V(γb2)≈ N−a22 . By the independence of bγ1 andbγ2 (see Theorem 2.4.1, Khuri et al., 1998) we have V

γb =diag2 1

a−1,N−a22

V−1/2

γb (γb−γ)→d N(0,I2). (1)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 6 / 32

(19)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2

⇒ eγ1 =bγ1 andeγ2=bγ2 ⇒`11= 0.

In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(20)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2

⇒ eγ1 =bγ1 andeγ2=bγ2 ⇒`11= 0.

In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(21)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2

⇒ γe1 =bγ1 andeγ2=bγ2 ⇒`11= 0. In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(22)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2

⇒`11= 0. In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(23)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2 ⇒`11= 0.

In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(24)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2 ⇒`11= 0.

In case γ1 ≤γ2

⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(25)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2 ⇒`11= 0.

In case γ1 ≤γ2 ⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(26)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2 ⇒`11= 0.

In case γ1 ≤γ2 ⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(27)

One-way model

Likelihood ratio

Using the Self and Liang approach (Self and Liang, 1987) and (1)

`1=−a−1 2

1

γ1

−1 2

−N−a 2

γb2

γ2

−1 2

In case γ1 > γ2 ⇒ γe1 =bγ1 andeγ2=γb2 ⇒`11= 0.

In case γ1 ≤γ2 ⇒ γ12

`12=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

d`12

−1 =−bγ1(a−1)

1

γ −1

−bγ2(N−a)

2

γ −1

= 0.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 7 / 32

(28)

One-way model

Likelihood ratio

γ1≤γ2

eγ = (a−1)bγ12+ (N−a)bγ22 (a−1)bγ1+ (N−a)bγ2

, `12=−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

Under the null hypothesis

`0=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

eγ = (a−1)γb12+ (N−a)bγ22

(a−1)γb1+ (N−a)bγ2 `0 =− (a−1) (N−a) (bγ1−bγ2)2 2 (a−1)bγ12+ (N−a)bγ22

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 8 / 32

(29)

One-way model

Likelihood ratio

γ1≤γ2

eγ = (a−1)bγ12+ (N−a)bγ22 (a−1)bγ1+ (N−a)bγ2

, `12=−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

Under the null hypothesis

`0=−a−1 2

1

γ −1 2

−N−a 2

2

γ −1 2

eγ = (a−1)γb12+ (N−a)bγ22

(a−1)γb1+ (N−a)bγ2 `0 =− (a−1) (N−a) (bγ1−bγ2)2 2 (a−1)bγ12+ (N−a)bγ22

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 8 / 32

(30)

One-way model

Likelihood ratio

γ1≤γ2

eγ = (a−1)bγ12+ (N−a)bγ22 (a−1)bγ1+ (N−a)bγ2

, `12=−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

Under the null hypothesis

`0=−a−1 2

1

γ −1 2

−N−a 2

γb2

γ −1 2

eγ = (a−1)γb12+ (N−a)bγ22

(a−1)bγ1+ (N−a)bγ2 `0 =− (a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)bγ22

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 8 / 32

(31)

One-way model

Likelihood ratio

γ1≤γ2

eγ = (a−1)bγ12+ (N−a)bγ22 (a−1)bγ1+ (N−a)bγ2

, `12=−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

Under the null hypothesis

`0=−a−1 2

1

γ −1 2

−N−a 2

γb2

γ −1 2

eγ = (a−1)γb12+ (N−a)bγ22

(a−1)bγ1+ (N−a)bγ2 `0 =− (a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)bγ22

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 8 / 32

(32)

One-way model

Likelihood ratio

γ1≤γ2

eγ = (a−1)bγ12+ (N−a)bγ22 (a−1)bγ1+ (N−a)bγ2

, `12=−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

Under the null hypothesis

`0=−a−1 2

1

γ −1 2

−N−a 2

γb2

γ −1 2

eγ = (a−1)γb12+ (N−a)bγ22

(a−1)bγ1+ (N−a)bγ2 `0 =−(a−1) (N−a) (bγ1−γb2)2 2 (a−1)bγ12+ (N−a)γb22

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 8 / 32

(33)

One-way model

Likelihood ratio

Proposition

The LRT rejects the null hypothesis for large values of λgiven by:

λ=

0 bγ1≤bγ2

(a−1)(N−a)(bγ1bγ2)2

((a−1)bγ12+(N−a)bγ22) bγ1>bγ2.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 9 / 32

(34)

One-way model

Random permutations

y = [yij] – lexicographical order

i – first factor j – replicates Permutation

yl =Ply, where Pl is a random permutation matrix.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 10 / 32

(35)

One-way model

Permutation procedure

1 Calculate ˆγ1 and ˆγ2

2 Calculateλ

3 Generate a pseudo-sample of size M of

`l =`(yl) =`(Ply)

4 Retrieve cα, the pseudo-sample’s empirical (1−α)-th quantile

5 Compare with the observed value λwith cα

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 11 / 32

(36)

One-way model

Application of delta method

Combining the delta method (see van der Vaart, 1998) with the Self and Liang approach (Self and Liang, 1987) we obtain the following proposition.

Proposition

The LRT rejects the null hypothesis H0 for a large values of κgiven by

κ=

(0 bγ1 ≤bγ2,

(a−1)(N−a)(logbγ1−logbγ2)2

2(N−1)1 >bγ2.

Theorem

The distribution of the LRT κ is a mixture of χ20 andχ21 with coefficients p and 1-p, with

p =P(bγ1 ≤bγ2).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 12 / 32

(37)

One-way model

Application of delta method

Combining the delta method (see van der Vaart, 1998) with the Self and Liang approach (Self and Liang, 1987) we obtain the following proposition.

Proposition

The LRT rejects the null hypothesis H0 for a large values ofκ given by

κ=

(0 γb1 ≤bγ2,

(a−1)(N−a)(logbγ1−logbγ2)2

2(N−1) γb1 >bγ2.

Theorem

The distribution of the LRT κ is a mixture of χ20 andχ21 with coefficients p and 1-p, with

p =P(bγ1 ≤bγ2).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 12 / 32

(38)

One-way model

Application of delta method

Combining the delta method (see van der Vaart, 1998) with the Self and Liang approach (Self and Liang, 1987) we obtain the following proposition.

Proposition

The LRT rejects the null hypothesis H0 for a large values ofκ given by

κ=

(0 γb1 ≤bγ2,

(a−1)(N−a)(logbγ1−logbγ2)2

2(N−1) γb1 >bγ2.

Theorem

The distribution of the LRT κ is a mixture of χ20 andχ21 with coefficients p and 1-p, with

p =P(bγ1 ≤bγ2).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 12 / 32

(39)

Two-fold nested model

Statistical model

y =µ1N+Z1α+Z2β+e, where

y - vector of observed values, µ- general mean,

1N - vector of one’s,

α andβ- vectors of random effects,

Z1 =Ia⊗1b⊗1r andZ2 =Ia⊗Ib⊗1r - are design matrices, e - vector of errors

We assume that α,β ande are indendent and thatα∼N 0, σa2Ia , β∼N 0, σb2Ib

ande ∼N 0, σe2IN

,respectively.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 13 / 32

(40)

Two-fold neste model

Statistical model

The problem of interest is:

H02b= 0 vs. H12b>0.

Alternatively, the problem of interest can be written as: H0023 vs. H102 > γ3,

whereγ1 =brσa2+rσ2be22 =rσ2b2e andγ32e, respectively (see Khuri et al, 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 14 / 32

(41)

Two-fold neste model

Statistical model

The problem of interest is:

H02b= 0 vs. H12b>0.

Alternatively, the problem of interest can be written as:

H0023 vs. H102 > γ3,

whereγ1=brσa2+rσ2b2e2 =rσ2b2e andγ32e, respectively (see Khuri et al, 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 14 / 32

(42)

Two-fold neste model

Statistical model

The problem of interest is:

H02b= 0 vs. H12b>0.

Alternatively, the problem of interest can be written as:

H0023 vs. H102 > γ3, whereγ1=brσa2+rσ2b2e2 =rσ2b2e andγ32e, respectively (see Khuri et al, 1998).

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 14 / 32

(43)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;

E(bγ2) =γ2 and;V (bγ2)≈ ab22; E(bγ3) =γ3 and; V(bγ3)≈ abr23.

b =diag2 1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(44)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;

E(bγ2) =γ2 and;V (bγ2)≈ ab22; E(bγ3) =γ3 and; V(bγ3)≈ abr23.

b =diag2 1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(45)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;E(bγ2) =γ2 and;V (bγ2)≈ ab22;

E(bγ3) =γ3 and; V(bγ3)≈ abr23. Vγb =diag2

1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(46)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;E(bγ2) =γ2 and;V (bγ2)≈ ab22; E(bγ3) =γ3 and;V (bγ3)≈ abr23.

b =diag2 1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(47)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;E(bγ2) =γ2 and;V (bγ2)≈ ab22; E(bγ3) =γ3 and;V (bγ3)≈ abr23.

b =diag2 1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(48)

Two-fold nested model

Statistical model

The maximum likelihood estimators (MLE)

µb= 1

N10Ny, bγ1 = y0P1y

a−1 , bγ2= yP2y

a(b−1), bγ3 = y0P3y ab(r−1), whereP1= Ia1aJa

br1Jbr,P2=Ia⊗ Ib1bJb

1rJr, P3=Iab⊗ Ir1rJr

.

E(bγ1) =γ1 andV(bγ1)≈ a12;E(bγ2) =γ2 and;V (bγ2)≈ ab22; E(bγ3) =γ3 and;V (bγ3)≈ abr23.

b =diag2 1

a ,ab22,abr23

V−1/2

γb (γb−γ)→d N(0,I3). (2)

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 15 / 32

(49)

Two-fold nested model

Likelihood ratio. H1

1 ≥γb2 ≥bγ3

1 =bγ12 =bγ23 =bγ3

−2`1 = 0

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 16 / 32

(50)

Likelihood Ratio

H1

1 <bγ2,bγ2 ≥bγ3

12= bγ21+bγb221+bγb2

3 =bγ3

−2`1 = ab2(bγ1γb2−bγ22)2+ab(bγ12−bγ12)2 (bγ12+bγb22)2

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 17 / 32

(51)

Two-fold nested model

Likelihood ratio. H1

1bγ22+rbγ32

bγ2+rbγ3,bγ2 <bγ3

1 =bγ1

23= bγ22+rbγ322+rbγ3

−2`1= abr2(bγ2γb3−bγ23)2+abr(bγ23−bγ22)2 (bγ22+rbγ32)2

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 18 / 32

(52)

Two-fold nested model

Likelihood ratio. H1

1 < bγ22+rbγ32

bγ2+rbγ3,bγ2 <bγ3

γe123= bγ12+bbγ22+brbγ321+bbγ2+brbγ3

−2`1 = a b(bγ12−bγ22) +br(γb13−bγ32)2

(bγ12+bγb22+brbγ32)2 +ab (bγ12−bγ12) +br(γb23−bγ32)2

(bγ12+bbγ22+brγb32)2

+abr (bγ12−bγ12) +b(γb12−bγ22)2

(bγ12+bbγ22+brγb32)2

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 19 / 32

(53)

Two-fold nested model

Likelihood ratio. H023

1bγ22+rbγ32

bγ2+rbγ3

1 =bγ1

23= bγ22+rbγ322+rbγ3

−2`1= abr2(bγ2γb3−bγ23)2+abr(bγ23−bγ22)2 (bγ22+rbγ32)2

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 20 / 32

(54)

Two-fold nested model

Likelihood ratio. H023

1 < bγ22+rbγ32

bγ2+rbγ3

γe123= bγ12+bbγ22+brbγ321+bbγ2+brbγ3

−2`0 = a b(bγ12−bγ22) +br(γb13−bγ32)2

(bγ12+bγb22+brbγ32)2 +ab (bγ12−bγ12) +br(γb23−bγ32)2

(bγ12+bbγ22+brγb32)2

+abr (bγ12−bγ12) +b(γb12−bγ22)2

(bγ12+bbγ22+brγb32)2

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 21 / 32

(55)

Two-fold nested model

Random Permutations

y = [yijk] – lexicographical order

i – first factor j – second factor k – replicates Selective permutation

yl = (Ia⊗Pl)y, where Pl is a random permutation matrix.

M. Przystalski, M. Fonseca Hypothesis testing in variance components with constraints LinStat 2014 22 / 32

Referências

Documentos relacionados

A pós-graduação não é uma atividade, tipicamente, universitária - o processo todo se elabora fora dos centros de decisão da universidade, embora o órgão controlador use