ANALYSING REGIONAL EXPORT DATA BY THE MODIFIED GENERALIZED F-TEST
Year 2018,
18. EYI Special Issue, 541 - 552, 20.01.2018
Mustafa Çavuş
,
Berna Yazıcı
,
Ahmet Sezer
Abstract
Classical
F-test is used for testing equality of more than two group means under
normality and variance homogeneity. Classical F test is most powerful
parametric method among the parametric statistical methods in case of the
assumptions are hold. However, the assumptions are not always satisfied in real
life. Thus researchers study on improving methods to solve this problem. Welch,
Generalized F, Parametric Bootsrap tests are proposed for testing equality of
group means under variance heterogeneity. These methods just give better
results under variance heterogeneity but they are not same in case of violation
of normality assumption due to researches. In this article, modified generalized
F-test is considered which is proposed for variance heterogeneity and
non-normality caused by outlier. To show the efficiency of this method, testing
equality of annual export amounts of geographical regions under variance
heterogeneity and non-normality caused by outlier. As a result, it is stated
that significant differences between regions are detected only by modified
generalized F-test.
References
- Alvandi, S. M., Jafari, A. A. and Mardani, F. A. (2012). One-way ANOVA with unequal variances. Communications in Statistics, 41, 4200-4221.
- Çavuş, M., Yazıcı, B. and Sezer, A. (2017). Modified tests for comparison of group means under heteroskedasticity and non-normality caused by outlier(s). Hacettepe Journal of Mathe- matics and Statistics, 46(3), 492-510.
- Çavuş, M., Yazıcı, B. and Sezer, A. (2017). Comparison of the average amounts of exports for geographical regions in Turkey with the modified generalized F-test. 18.International SYmpo- sium on Econometrics Operations Research and Statistics, Trabzon.
- Fisher, R. A. (1925). Statistical methods and scientific inference. Oxford, England: Oliver and Boyd.
- Gamage, J. and Weerahandi, S. (1998). Size performance of some tests in one-way ANOVA. Communications in Statistics, 27(3), 625-640.
- Huber, P. J. (1964). Robust estimation of a location parameter. The Annals of Mathematical Statistics, 69, 73-101.
- Karagoz, D. (2015). Modified Welch test statistic for ANOVA under Weibull distribution. Hacettepe Journal of Mathematics and Statistics, 27(3), 625-640.
- Karagoz, D. and Saraçbaşı, T. (2009). Robust Brown-Forsythe and robust modified Brown-Forsythe ANOVA tests under heteroscedasticity for contaminated Weibull distribution. Revista Colombiana de Estadistica, 39, 17-32.
- Krishnamoorthy, K., Lu, F. and Mathew, T. (2007). A parametric bootstrap approach for ANOVA with unequal variances. Computational Statistics and Data Analysis, 51, 5731-5742.
- Li, X. (2009). A generalized p-value approach for comparing the means of several log-normal populations. Statistics and Probability Letters, 79, 1404-1408.
- Tan, W. Y. and Tabatabai, M. A. (1985). Some robust ANOVA procedures under heteroscedasticity and non-normality. Communications in Statistics, 14(4), 1007-1026.
- Tsui, K. and Weerahandi, S. (1989). Generalized p-values in significance testing of hypotheses in the presence of nuisance parameters. Journal of the American Statistical Association, 84, 602-607.
- Welch, B. L. (1951). On the comparison of several group mean values. Biometrika, 38, 330-336.
- Weerahandi, S. (1995). ANOVA under unequal error variances. Biometrics, 51, 589-599.
- Wilcox, R. R. (1995). Simulation results on solutions to the multivariate Behrens-Fisher problem via trimmed means. The Statistician, 44(2), 213-225.
BÖLGESEL İHRACAT VERİLERİNİN MODİFİYE EDİLMİŞ GENELLEŞTİRİLMİŞ F-TESTİ İLE ANALİZİ
Year 2018,
18. EYI Special Issue, 541 - 552, 20.01.2018
Mustafa Çavuş
,
Berna Yazıcı
,
Ahmet Sezer
Abstract
Normal
dağılmış anakütlelerden geldiği bilinen homojen varyanslı ikiden fazla grubun
ortalamasının eşitliğinin test edilmesi için Klasik F-testi kullanılır. Klasik
F-testi, grupların birbirinden bağımsız, homojen varyanslı ve normal dağıldığı
varsayımları altında parametrik istatistiksel yöntemler arasında en güçlü testtir.
Gerçek hayatta bahsedilen varsayımların sağlandığı durumlarla çok nadirdir. Bu
nedenle araştırmacılar varsayımların sağlanmadığı durumlar için yöntemler
geliştirmeye yönelmişlerdir. Welch, Genelleştirilmiş F, Parametrik Bootstrap
testleri varyans homojenliğinin sağlanmadığı durumlarda normal dağılmış
grupların ortalamalarının eşitliğinin test edilmesi için geliştirilmiştir.
Yalnızca varyans homojenliği sağlanmadığı durumlarda doğru sonuçlar veren bu
yöntemler normal dağılım varsayımının bozulması durumunda performanslarını
kaybettiklerinden birçok çalışmada bahseilmiştir. Bu çalışmada aykırı değerden
kaynaklı normal dağılmama ve homojen olmayan varyanslılık durumunda grup
ortalamalarının karşılaştırılabilmesi için kullanılabilen Modifiye Edilmiş
Genelleştirilmiş F-testi ele alınmıştır. Bahsedilen koşullar altında bu yöntemin
etkinliğinin ortaya konulabilmesi için homojen varyansa sahip olmayan ve aykırı
değerden kaynaklı normal dağılmayan Türkiye’deki coğrafi bölgelerin ortalama
ihracat tutarları karşılaştırılmıştır. Sonuç olarak bölgeler arasındaki
istatistiksel olarak anlamlı farkların Modifiye edilmiş Genelleştirilmiş
F-testi ile tespit edilebileceği ortaya konulmuştur.
References
- Alvandi, S. M., Jafari, A. A. and Mardani, F. A. (2012). One-way ANOVA with unequal variances. Communications in Statistics, 41, 4200-4221.
- Çavuş, M., Yazıcı, B. and Sezer, A. (2017). Modified tests for comparison of group means under heteroskedasticity and non-normality caused by outlier(s). Hacettepe Journal of Mathe- matics and Statistics, 46(3), 492-510.
- Çavuş, M., Yazıcı, B. and Sezer, A. (2017). Comparison of the average amounts of exports for geographical regions in Turkey with the modified generalized F-test. 18.International SYmpo- sium on Econometrics Operations Research and Statistics, Trabzon.
- Fisher, R. A. (1925). Statistical methods and scientific inference. Oxford, England: Oliver and Boyd.
- Gamage, J. and Weerahandi, S. (1998). Size performance of some tests in one-way ANOVA. Communications in Statistics, 27(3), 625-640.
- Huber, P. J. (1964). Robust estimation of a location parameter. The Annals of Mathematical Statistics, 69, 73-101.
- Karagoz, D. (2015). Modified Welch test statistic for ANOVA under Weibull distribution. Hacettepe Journal of Mathematics and Statistics, 27(3), 625-640.
- Karagoz, D. and Saraçbaşı, T. (2009). Robust Brown-Forsythe and robust modified Brown-Forsythe ANOVA tests under heteroscedasticity for contaminated Weibull distribution. Revista Colombiana de Estadistica, 39, 17-32.
- Krishnamoorthy, K., Lu, F. and Mathew, T. (2007). A parametric bootstrap approach for ANOVA with unequal variances. Computational Statistics and Data Analysis, 51, 5731-5742.
- Li, X. (2009). A generalized p-value approach for comparing the means of several log-normal populations. Statistics and Probability Letters, 79, 1404-1408.
- Tan, W. Y. and Tabatabai, M. A. (1985). Some robust ANOVA procedures under heteroscedasticity and non-normality. Communications in Statistics, 14(4), 1007-1026.
- Tsui, K. and Weerahandi, S. (1989). Generalized p-values in significance testing of hypotheses in the presence of nuisance parameters. Journal of the American Statistical Association, 84, 602-607.
- Welch, B. L. (1951). On the comparison of several group mean values. Biometrika, 38, 330-336.
- Weerahandi, S. (1995). ANOVA under unequal error variances. Biometrics, 51, 589-599.
- Wilcox, R. R. (1995). Simulation results on solutions to the multivariate Behrens-Fisher problem via trimmed means. The Statistician, 44(2), 213-225.