An Alternative Solution to Hotelling T square under the Heteroscedasticity of the dispersion Matrix

P.O. Adebayo* and G.M. Oyeyemi

Department of Statistics, University of Ilorin, Ilorin, Nigeria

*Corresponding Author:
P.O. Adebayo
Department of Statistics, University of Ilorin, Ilorin, Nigeria.
E-mail:
bayooni3@gmail.com

Received Date: 12/11/2018; Accepted Date: 23/11/2018; Published Date: 28/11/2018

Visit for more related articles at Research & Reviews: Journal of Statistics and Mathematical Sciences

Abstract

This work focus on developing an alternative procedure to the multivariate Behrens-Fisher problem by using an approximate degree of freedom test which was adopted from Satterthwaite univariate procedure. The proposed procedure was compared via R package simulation and real-life data used by Tim (1975) with six (6) existing procedures namely: Johanson, Yao, Krishnamoorthy, Hotelling T square, Nel and Van der Merwe and Yanagihara. And it was discovered that proposed procedure performed better in term of power of the test than all existing procedures considered in all the scenarios that are at different: (i) random variables (p), (ii) variancecovariance matrix, (iii) sample size and (iv) significant level (α). And compete favorably well in term of type I error rate with Johanson, Yao, Krishnamoorthy, Nel, and Van der Merwe.

Keywords

Variance co-variance matrix, linear combination, Type I error rate, power the test, Heteroscedasticity, R statistical package

Introduction

Suppose we have a random sample of size n1, x11, x12, x13, ……x1n1 for N(μ112) and a second random sample of size n2, x21, x22, x23, . . . x2n2 for N(μ222 ). It is desired to test H0: u1= u2 against H1: u1≠u2. If σ1 and σ2 are both known a normal test is used. If σ12 but both are unknown a t-test is commonly used with the test statistics.

Equation

Where Equation

Equation

A t-test is a type of statistical test that is used to compare the means of two groups, such as men vs. women, athletes vs. non-athletes, young vs. elderly, or you may want to compare means measured on a single group under two different experimental conditions or at two different times. T-tests are a type of parametric method; they can be used when the samples satisfy the conditions of normality, equal variance, and independence. T-tests can be divided into two types. There is the independent t-test, which can be used when the two groups under comparison are independent of each other, and the paired t-test, which can be used when the two groups under comparison are dependent on each other. If σ1≠σ2 and both are unknown then we are confronted with the Behrens-Fisher problem. There is no universally accepted testing procedure for this problem although arrays of tests have been developed and will be discussed in the Review of Literature. Behrens [1] proposed the statistics.

In the literature, there is a modified test statistic (t-test) when the assumption of equal variances is violated has been known as the Behrens-Fisher problem [1,2]. Early investigations showed that the problem can be overcome by substituting separatevariances tests, such as the ones introduced by Welch [3,4], and Satterthwaite [5], for the Student t-test. These modified significance tests, unlike the usual two-sample Student t-test, do not pool variances in the computation of an error term. Moreover, they alter the degrees of freedom by a function that depends on sample data. It has been found that these procedures in many cases restore Type I error probabilities to the nominal significance level and also counteract increase or decrease of Type II error [6-9].

Student’s t-test is univariate and analog to Hotelling T square which is the multivariate version of T-test and this Hotelling’s T2 has three basic assumptions that are fundamental to the statistical theory: independent, multivariate normality and equality of variance-covariance matrices. A statistical test procedure is said to be robust or insensitive if departures from these assumptions do not greatly affect the significance level or power of the test.

To use Hotelling’s T2 one must assume that the two samples are independent and that their variance-covariance matrices are equal (Σ12=Σ). When variance-covariance matrices are not homogeneous and unknown, the test statistic will not be distributed as a T2. This predicament is known as the multivariate Behrens-Fisher problem.

The Behrens-Fisher Problem addresses interval estimation and hypothesis testing concerning the differences between the means of two normally distributed populations when the variances of the two populations are not equal. While Multivariate Behrens- Fisher problem deal with testing the equality of two normal mean vectors under heteroscedasticity of dispersion matrices. These are the some of the existing Multivariate Behrens-Fisher problem: Yao [10], Johansen [11], Nel et al [12], Kim [13], Krishnamoorthy and Yu [14], Gamage et al [15] Yanagihara and Yuan [16],and Kawasaki and Seo [17] and so on. But with all these procedures there is no one with a hundred percent (100%) satisfactory in term of power of the test and type I error rate. And each of these scholars works on the degree of freedom using a different method which is classified into four (4): Approximate degree of freedom tests, Series expansion-based tests, Simulation-based tests, and Transformation-based tests.

Scheffe [18],Lauer and Han [19], Lee and Gurland [20], Murphy [21], Yao [10], Algina and Tang [22], Kim [13], De la Rey and Nel’s [23], Christensen and Rencher [24] , Oyeyemi and Adebayo [25]. All these authors mentioned and many more have work on the comparison of some of the Multivariate Behrens-Fisher problem procedures.

The purpose of this work is to develop an alternative procedure for multivariate data that will be robust compared to other procedures and the work will begin with an introduction to the statistical notation that will be helpful in understanding the concepts. This is followed by a discussion of procedures that can be used to test the hypothesis of multivariate mean equality when statistical assumptions are and are not satisfied. We will then show how to obtain a test that is robust to the covariance heterogeneity.

Multivariat Behrens-Fisher Problem (Existing Procedure)

Consider two p-variate normal populations Equation andEquation where Equation and Equation are unknown p × 1 vector and Σ1 and Σ2 are unknown p × p positive definite matrices.

Let Equation and

Equation denote random samples from these two populations, respectively. We are interested in the testing problem

Equation

For i=1,2, let

Equation

Equation

Equation

Then Equation andEquation which are sufficient for the mean vectors and dispersion matrices, are independent random variables having the distributions:

Equation

Where Wp(r,Σ)denotes the p -dimensional wishart distribution with df=r and scale matrix Σ.

Equation

Equation

Equation

The following are the existing procedures or solutions to Multivariate Behrens-Fisher problem considered in this study

Yao Procedure

Yao [10], invariant test. This is a multivariate extension of the Welch ‘approximate degree of freedom’ solution provided by Turkey and the test statistic is based on a transformation of Hotelling T2. And is based on Equation with the

Equation

Equation

Statistical significance is then assessed by comparing the TYao statistic to its critical value Fα(p,v-p+1), that is, a critical value from the F distribution with p and v–p+1 degrees of freedom (df)

Johansen Procedure

Johansen [11], invariant test, Yanagihara and Yuan [16], Kawasaki and Takashi [17]. They used T2 ~ qFpv where

Equation

Equation

Equation

And his proposed test statistic Equation

Statistical significance is then assessed by comparing the TJohan statistic to its critical value Fα(p,vJoh), that is, a critical value from the F distribution with p and vJoh degrees of freedom (df).

Nel and Van der Merwe (1986) noninvariant solution

Here we use

Equation except that v is defined by

Equation

Krishnamoorthy and Yu (2004)’s, Lin and Wang (2009), Modified Nel/ Van der Merwe Invariant Solution

We use the idea as before, namely,

Equation with the d.f.v defined by

Equation

Equation

Equation

Statistical significance is then assessed by comparing the TKrish statistic to its critical value Fα (p,v_kY-p+1), that is, a critical value from the F distribution with p and v_kY-p+1 degrees of freedom (df)

Yanagihara and Yuan Procedure used the Series Expansion Based Test to Developed an Alternative Procedure to Multivariate Behrens-Fisher problem

Equation

Equation

Equation

Equation

Equation

Equation

Hotelling’s T2

Equation

Where

Equation

The test statistic can also be converted to an F statistic,

Equation

Where N=n1+n2. Statistical significance is then assessed by comparing the vhotel statistic to its critical value Fα (p, N-p-1), that is, a critical value from the F distribution with p and N-p-1 degrees of freedom (df).

The Mean and the Variance of the Χ2 Distribution with N degrees of Freedom

The chi-square distribution is defined with n degrees of freedom by

Equation whereEquation are independent random variables, each with distribution N(0,1).

Find the expected value and variance of both sides, then we have

Equation and

Equation

And all the instances of Zi have identical distributions, then

Equation andEquation

Where Z is the random variable with distribution N(0, 1). Then

Equation

Therefore

Equation (1)

For Var(Z2)

Equation

Equation

Now

Equation (2)

To find E(Z4), we will use the fact that for any continuous random variable X with probability density function f, and any exponent k,

Equation

And that the probability density function f of the N(0,1) random variable is given by

Equation

Then, Equation

By integration by parts, we have, Equation

Equation

Equation (3)

Therefore substitute equation (3) into equation (2) then we have

Equation (4)

For two sample t-test, we will limit this work to the version of the test where we do not assume that the two populations have equal variances. Let random sample x1,…,xn1 from a random variable X with distribution N(μ11) and a random sample y1,….,yn2 from a random variable Y with distribution (μ22). We have

Equation Strictly speaking, this statistic does not follow t-distribution, therefore;

The variance of Equation isEquation and as an estimator forEquation we haveEquation

For t to bet–distribution, there would have to be some multiple of Equation that is chi-square distribution and this is not the case. However, remember that in the one-sample case, Equation had a chi-square distribution with n-1 degrees of freedom. By analogy, we have Equation has a chi-square distribution with r degrees of freedom. Satterthwaite found the true distribution of Equation and showed that if r is chosen so that the variance of the chi-square distribution with r degrees of freedom is equal to the true variance of Equation, then, under certain conditions, this chi-square distribution with r degrees of freedom is a good approximation to be the true distribution of Equation so from this point, we are assuming thatEquation has distributionEquation. So from equation                                    (4)

Equation (5)

Equation (6)

Equating the equation (5) and (6)

Equation

Equation (7)

Now Equation and s1 and s2 are independent so

Equation (8)

We know that Equation has a chi-square distribution with n1-1 degrees of freedom, from equation (3)

Equation

Equation

Equation (9)

Substitute equation (9) into equation (8)

Equation (10)

Substitute equation (10) into equation (7)

Equation

Equation (11)

In practice, the values of the population variances, Equation andEquation are unknown, and so we replaceEquation and Equationby their estimators Equation and Equation alsoEquation from equation (11)

Equation

Multivariate Approach

We shall consider the test statistic y'S-1 y and use Univariate Satterthwaite approximation of degrees of freedom method to suggest multivariate generalization based on the T2–distribution. We have

Equation

Equation

Equation

Where b is an arbitrary constant vector

Equation

Equation Multivariate of a version of equation (5)

Equation (12)

Multivariate of a version of equation (5) is equation (12)

Equation

Equation

Equation

Equation

Note Equation (14)

Put equation (14) into equation (13)

Equation

Equation (15)

Multivariate of a version of equation (9) is

Equation (16)

Put equation (16) into equation (15)

Equation (17)

Equation (17) becomes

Equation

The values of the population variances Σi are unknown, and so we replace Σi and b'Σb by their estimators Si and b'Sb

Equation (18)

Put equation (14) into equation (18) to have

Equation (19)

Set b=S-1 y then equation (19) becomes

Equation (20)

Equation (20) can be in this form

Equation (21)

Let Equation then equation (22) become

Equation

Equation (22)

Equation (23)

Then equation (23) is the test statistic of the proposed procedure. Statistical significance is then assessed by comparing the TProp statistic to its critical value Equation that is, a critical value from the F distribution with p and Equation degrees of freedom (df)

Simulation Study

A simulation study using R package was conducted in order to estimate and compare the Type I error rate and power for each of the previously discussed approximate solution [10,11,14], Proposed procedure, Hotelling’s T square, [14,16]. The simulations are carried out when the null hypothesis is true and not true, for Multivariate normal distribution, when there are unequal variancecovariance matrix. Five (5) factors were varied in the simulation: the sample size, the number of variables p, variance covariance matrices, mean vectors, and significant levels. For each of the above combinations, an ni ×p data matrix Xi(i=1 and 2) were replicated 1,000. The comparison criteria; type I error rate and power of the test were therefore obtained and the results were presented in both tabular.

The following are the levels used for each of the three factors.

These levels provide 36-factor combinations the values for sample size are shown in Table 1.

Multivariate Distribution P a Sample size
Normal 2, 3, 4 0.01 20, 10
2, 3, 4 0.025 50, 30
2, 3, 4 0.05 100, 60

Table 1: Levels used for each of the three factors.

Result

From the Table 2 Nel and Van der Merwe have the highest power of the test when the sample sizes are small (20, 10) but at (50, 30) and (100, 60) the proposed procedure has the highest power than all other procedures.

P=2  a=0.01
X1=(20 30)  n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.3852 0.3872 0.3872 0.3861 0.2146 0.3915 0.2487
X2=(10 30)  50, 30 0.8332 0.8349 0.8346 0.8374 0.6672 0.8348 0.7691
100, 60 0.991 0.9911 0.9911 0.9912 0.9497 0.9911 0.9865
 Equation  a=0.025
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.5037 0.5043 0.5045 0.5037 0.3158 0.508 0.3787
50, 30 0.8903 0.8911 0.8909 0.8926 0.7609 0.891 0.8528
Equation 100, 60 0.9959 0.9959 0.9959 0.996 0.973 0.9959 0.9941
 a=0.05
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.6106 0.6105 0.6104 0.6099 0.4201 0.6133 0.5092
50, 30 0.9319 0.932 0.9319 0.9329 0.8348 0.932 0.911
100, 60 0.9982 0.9982 0.9982 0.9983 0.9855 0.9982 0.9975

Table 2: Power of the test.

From Table 3, when the sample size is (20,10) the proposed procedure is on nominal level exactly while Hoteling T square and Yanagihara [16] are below the nominal level, but at (50,30) and (100,60) all the procedures are below the nominal level, at significant level 0.01. At α=0.025, all the procedures are inflated at (50,30) and deflated at (20,10) and (100,60).

P=2  a=0.01
X1=(20 30)  n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.011 0.011 0.011 0.01 0.002 0.011 0.002
 X2=(20 30) 50, 30 0.007 0.007 0.007 0.007 0.001 0.006 0.004
100, 60 0.008 0.008 0.008 0.008 0 0.008 0.008
 Equation  a=0.025
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.017 0.017 0.017 0.018 0.003 0.018 0.005
50, 30 0.026 0.027 0.026 0.027 0.008 0.026 0.015
Equation 100, 60 0.015 0.015 0.015 0.015 0.002 0.015 0.009
 a=0.05
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.048 0.049 0.049 0.048 0.006 0.049 0.023
50, 30 0.045 0.044 0.045 0.046 0.011 0.045 0.035
100, 60 0.057 0.057 0.057 0.057 0.016 0.057 0.051

Table 3: Type I error rate.

From Table 4, It is obvious that proposed procedure performed better than all other procedures at (50,30) and (100, 60) but Nel and Van der Merwe is better when the sample size is (20,10) in all the scenarios considered.

P=3  a=0.01
Equation n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.2029 0.2198 0.2054 0.2147 0.1313 0.2278 0.0895
Equation 50, 30 0.6107 0.6182 0.6154 0.6217 0.4472 0.6203 0.5187
100, 60 0.9295 0.9312 0.9307 0.9322 0.8107 0.9315 0.9073
   Equation  a=0.025
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.2954 0.312 0.2983 0.3062 0.206 0.3205 0.1684
50, 30 0.7137 0.7195 0.7173 0.7222 0.5602 0.7209 0.6451
Equation 100, 60 0.9596 0.9606 0.9603 0.9611 0.877 0.9608 0.9478
 a=0.05
n1 ≠ n2 Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.4095 0.4255 0.4122 0.4207 0.3002 0.4334 0.2743
50, 30 0.7952 0.7992 0.7978 0.8012 0.6624 0.8003 0.7475
100, 60 0.9754 0.976 0.9758 0.9763 0.9199 0.9761 0.9692

Table 4: Power of the test.

Table 5, the type I error rate of all procedures considered are fluctuating, either inflated or deflated. At significant level 0.01, 0.02, 0.05 their inflation in type I error rate, when sample sizes are (50,30) and (100,60), but at (20, 10) all most all the procedures are deflated.

P=3  a=0.01
Equation   Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.007 0.008 0.008 0.006 0.014 0.013 0
Equation 50, 30 0.011 0.01 0.011 0.09 0.013 0.011 0.006
100, 60 0.015 0.015 0.015 0.015 0.018 0.015 0.013
   Equation  a=0.025
  Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.035 0.033 0.035 0.027 0.033 0.041 0.003
50, 30 0.025 0.027 0.026 0.024 0.032 0.028 0.011
Equation 100, 60 0.029 0.031 0.031 0.03 0.039 0.031 0.026
 a=0.05
  Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.059 0.066 0.068 0.048 0.049 0.073 0.013
50, 30 0.053 0.059 0.055 0.056 0.052 0.057 0.042
100, 60 0.061 0.061 0.062 0.061 0.056 0.063 0.051

Table 5: Type I error rate.

Table 6 shows when that sample size is (20,10) Nel and Van der Merwe performed better than other procedures, but when sample size increases to (50,30) proposed procedure is better. And there was a great competition among the procedures at (100,60).

P=4  a=0.01
Equation   Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.343 0.3589 0.3599 0.3508 0.1974 0.3737 0.078
 Equation 50, 30 0.86 0.8712 0.8711 0.875 0.7027 0.8734 0.6825
100, 60 0.9968 0.9971 0.9971 0.9973 0.9733 0.9972 0.9902
Equation  a=0.025
  Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.4656 0.48 0.4817 0.4721 0.297 0.4946 0.1604
50, 30 0.918 0.9249 0.9247 0.9271 0.799 0.926 0.8057
Equation 100, 60 0.9991 0.9992 0.9992 0.9992 0.9888 0.9992 0.9971
 a=0.05
  Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.5662 0.5793 0.5804 0.5731 0.3905 0.5914 0.2585
50, 30 0.9507 0.9547 0.9547 0.9561 0.8649 0.9555 0.886
100, 60 0.9995 0.9996 0.9996 0.9996 0.9938 0.9996 0.9987

Table 6: Type I error rate

Hotelling T [26] square and yanagihara [16] are below the nominal level in all the scenarios considered, while other procedures fluctuated (Inflated or deflated) round the nominal level as in Table 7.

P=4  a=0.01
Equation   Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.008 0.013 0.011 0.011 0.002 0.013 0
 Equation 50, 30 0.011 0.012 0.012 0.012 0.001 0.012 0.002
100, 60 0.007 0.007 0.007 0.007 0 0.007 0.004
Equation  a=0.025
  Johan Yao Krish Propo Hotel Nel Yana
20, 10 0.024 0.029 0.031 0.026 0.004 0.033 0
50, 30 0.02 0.026 0.026 0.026 0.003 0.026 0.006
Equation 100, 60 0.021 0.022 0.022 0.022 0.002 0.022 0.01
 a=0.05
  Johan Yao Krish Propo Hotel Nel Yana
Equation 20, 10 0.05 0.055 0.055 0.051 0.009 0.06 0
50, 30 0.048 0.051 0.051 0.051 0.012 0.051 0.022
100, 60 0.052 0.053 0.052 0.053 0.008 0.053 0.034

Table 7: Power of the test.

Data Set (For Illustrated Example)

The data used here is an illustrated example used by Timm. The two sample sizes considered are ten and twenty respectively (n1=10 and n2=20) and two random variables (p=2) form each population.

The sample means and their covariances are

Equation

Equation

n1=10, n2=20

The difference between the means is

Equation

And the test statistic is

Equation

Equation

T2=11.58542

From Table 8, the proposed procedure has the highest power followed by Yanagihara [16], Krishnamoorthy [14], Yao [10], Hotelling T square [26], Johansen [11] and Nel and Van der Merwe [14] at all the significant level α considered (α=0.05, 0.025 and 0.01).

   a=0.05
  Johan Yao Krish Propo Hotel Nel Yana
Critical value 6.978 7.2012 7.223 7.7396 6.9567 6.9601 10.0088
Power 0.4979 0.5109 0.5121 0.868 0.5068 0.4969 0.6244
   a=0.025
Critical value 8.852 9.1661 9.1987 9.9867 8.7984 8.8036 13.8708
Power 0.62 0.6334 0.6347 0.9325 0.6273 0.618 0.7527
   a=0.01
Critical value 11.4986 11.9613 12.0129 13.2753 11.3828 11.3908 20.3456
Power 0.7503 0.7625 0.7638 0.9732 0.7551 0.7456 0.868

Table 8: The result from the illustrated example.

Findings

From the simulation, it is obvious from Tables 1, 3 and 5 that when the sample size is very small (20,10) proposed procedure is not at his best, but when sample size increases to (50,30) and (100,60), the proposed procedure performed better than the all procedures considered. Nel and Van der Merwe [14] performed better when the sample size is very small (20,10) followed by Yao [10], Krishnamoorthy [14] and Proposed procedures in term of power of the test in all the scenarios considered.

In term of Type I error rate, proposed procedure compete favorably well with the other procedures selected for this study. Yao [10], Krishnamoorthy [14], Johansen [11], Nel and Van der Merwe [14] and the proposed procedures are fluctuating (Inflated and Deflated) around the nominal level while Hotelling T [26] square and Yanagihara [16] are below the nominal level.

References