Sains Malaysiana 45(10)(2016):
1565–1572
A Bayesian Approach to the One Way
ANOVA under Unequal Variance
(Pendekatan Bayesian kepada ANOVA Sehala di bawah Varians tak Sama)
NOPPAKUN TONGMOL1,
WUTTICHAI
SRISODAPHOL2*
& ANGKANA BOONYUED1
1Department
of Mathematics, Faculty of Science, Khon Kaen University, Khon
Kaen 40002
Thailand
2Department
of Statistics, Faculty of Science, Khon Kaen University, Khon
Kaen 40002
Thailand
Diserahkan:
8 Mei 2015/Diterima: 15 Februari 2016
ABSTRACT
This study involves testing
the equality of several normal means under unequal variances,
which is the setup of one-way analysis of variances (one-way
ANOVA).
Several tests are available in the literature, however, most
of them perform poorly in terms of type I error rate under unequal
variances. In fact, Type I errors can be highly inflated for
some of the commonly used tests, a serious issue that seems
to have been overlooked. Even though several tests have been
proposed to overcome the problem, most of them show difficulty
in calculation. Accordingly, the test for ANOVA with estimation of parameters using
Bayesian approach is proposed as an alternative to such tests.
The proposed test is compared with four existing tests such
as the original test, James’s test, Welch’s test and the parametric
bootstrap (PB)
test. Type I error rates and powers of the tests are evaluated
using Monte Carlo simulation. Our results indicated that the
performance of the proposed test is superior to the original
test and is comparable to James’s test, Welch’s test and the
PB test, controlling Type I error rate
quite well and showing high power of the test. Our study suggested
that the proposed test has high performance and should be used
as an alternative to the four existing tests due to its simple
formula.
Keywords: Bayesian approach;
power of the test; Type I error rate; unequal variance
ABSTRAK
Kajian ini melibatkan
ujian kesamaan dalam beberapa cara yang biasa di bawah varians
tak sama yang merupakan persediaan varians analisis sehala (ANOVA sehala).
Beberapa ujian telah sedia ada dalam penulisan ilmiah, walau bagaimanapun,
tidak menunjukkan keputusan memberangsangkan daripada segi kadar
ralat Jenis I di bawah varians tak sama. Malah, ralat Jenis
I boleh melambung tinggi bagi sesetengah ujian yang biasa digunakan,
suatu isu yang serius yang seolah-olah telah diabaikan. Walaupun
beberapa ujian telah dicadangkan untuk mengatasi masalah ini,
sebahagian besar menunjukkan kesukaran dalam pengiraan. Sehubungan
dengan itu, ujian bagi ANOVA dengan parameter anggaran menggunakan
pendekatan Bayesian dicadangkan sebagai alternatif kepada ujian
tersebut. Ujian yang dicadangkan dibandingkan dengan empat ujian
sedia ada seperti ujian asal, ujian James, ujian Welch dan ujian
butstrap berparameter (PB).
Kadar ralat Jenis I dan kuasa ujian dinilai menggunakan simulasi
Monte Carlo. Keputusan kajian kami menunjukkan bahawa prestasi
ujian cadangan itu lebih cemerlang berbanding ujian asal dan
setanding dengan ujian James, Welch dan PB,
mengawal kadar ralat Jenis I dengan baik dan menunjukkan kuasa
tinggi ujian tersebut. Kajian kami menyarankan bahawa ujian
cadangan mempunyai prestasi yang tinggi dan harus digunakan
sebagai suatu alternatif kepada empat ujian sedia ada kerana
formula yang mudah.
Kata kunci: Kadar ralat Jenis I; kuasa ujian; pendekatan Bayesian;
varians tak sama
RUJUKAN
Algina,
J., Oshima, T.C. & Lin, W. 1994. Type I error rates for
Welch’s test and James’ second-order test under nonnormality
and inequality of variance when there are two groups. Journal
of Educational and Behavioral Statistics 19: 275-291.
Box,
G.E.P. 1954. Some theorems on quadratic forms in the study of
analysis of variance problems. I. Effect of inequality of variance
in the one-way classification. Annals of Mathematical Statistics
25: 290-302.
James,
G.S. 1951. The comparison of several groups of observations
when the ratios of population variances are unknown. Biometrika
38: 324-329.
Krishnamoorthy,
K., Lu, F. & Mathew, T. 2007. A parametric bootstrap approach
for ANOVA with unequal variances: Fixed and random models. Computational
Statistics and Data Analysis 51: 5731-5742.
Robert,
C.P. 2007. The Bayesian Choice from Decision-Theoretic Foundations
to Computational Implementation. 2nd ed. New York: Springer
Verlag.
Seber,
G.A.F. 1977. Linear Regression Analysis. New York: John
Wiley and Sons.
Welch,
B.L. 1951. On the comparison of several mean values: An alternative
approach. Biometrika 38: 330-336.
Welch,
B.L. 1937. The significance of the difference between two means
when the population variance are unequal. Biometrika 29:
350-362.
*Pengarang untuk surat-menyurat;
email: wuttsr@kku.ac.th