آزمون نسبت درستنمایی علامتدار تعدیل یافته برای ضریب تغییرات یک جامعه گاوسی وارون

نویسنده
دانشگاه فسا
چکیده
در این مقاله مسئله آزمون فرض­های آماری دوطرفه برای پارامتر ضریب تغییرات از یک جامعه گاوسی وارون مورد بررسی می­گیرد. روشی که در اینجا استفاده می­شود، روش نسبت درستنمایی علامت­دار تعدیل­یافته (MSLR) است که یک نسخه بهبود یافته از روش نسبت درستنمایی علامت­دار کلاسیک می­باشد. پژوهش­های انجام گرفته نشان دهنده آن است که دقت این روش از مرتبه سوم است و این در حالی است که دقت روش کلاسیک از مرتبه یک می­باشد. در حقیقت این گونه روش­ها بر پایه تابع درستنمایی از مرتبه دقت بالاتر هستند. به علت این دقت بالا، این مقاله بر آن است که از این روش برای استنباط در مورد پارامتر ضریب تغییرات از یک جامعه گاوسی وارون استفاده نماید. فرمول­ها و محاسبات لازم برای بدست آوردن آماره آزمون MSLR ارائه شده است. از نظر عددی، عملکرد روش MSLR در مقابل روش­های کلاسیک از لحاظ نرخ خطای نوع اول تجربی و اندازه توان تجربی مورد مقایسه قرار گرفته ­است. نتایج مطالعات شبیه­سازی نشان می­دهند که نرخ خطای نوع اول تجربی روش MSLR حتی برای اندازه نمونه­های کوچک هم به نرخ خطای نوع اول اسمی نزدیک می­باشد در حالیکه روش­های کلاسیک تنها برای اندازه نمونه­های بزرگ قابل اعتماد هستند. مقایسه توان­های تجربی هم نشان می­دهد که در بعضی از وضعیت­ها، اندازه توان روش مذکور از سایر روش­ها بهتر است، با توجه به این که روش­های رقیب در کنترل خطای نوع اول عملکرد مناسبی ندارند به این معنی که نرخ خطای نوع اول آنها از نرخ خطای نوع اول اسمی دور است. در نهایت، محاسبات مربوط به تمام روشها در یک مثال واقعی فراهم شده و سپس مقاله نتیجه­ گیری می­شود.
کلیدواژه‌ها

عنوان مقاله English

Modified signed log-likelihood test for the coefficient of variation of an inverse Gaussian population

نویسنده English

Mohammad Reza Kazemi
Fasa University
چکیده English

In this paper, we consider the problem of two sided hypothesis testing for the parameter of coefficient of variation of an inverse Gaussian population. An approach used here is the modified signed log-likelihood ratio (MSLR) method which is the modification of traditional signed log-likelihood ratio test. Previous works show that this proposed method has third-order accuracy whereas the traditional approach has first-order one. Indeed, these methods are based on likelihood with a higher order of accuracy. For this reason, we are interested in using this method for inference about the parameter of coefficient of variation of an inverse Gaussian distribution. All necessary formulas for obtaining MSLR statistic are provided. Numerically, the performances of this method are compared with classical approaches, in terms of empirical type-I error rate and empirical test power. Simulation results show that the empirical type-I error rates of MSLR are close to nominal type-I error rate, even for small sample sizes whereas the traditional approaches are reliable only for large sample sizes. Comparing the empirical power sizes shows that the power of MSLR method is superior to other considered methods in some settings, by regarding that the competing approaches cannot perform well in controlling the type-I error probability because their empirical type-I error rates are far from the nominal type-I error rate. Finally, we illustrate the proposed methods using a real data set and then we conclude the paper.

کلیدواژه‌ها English

Coefficient of Variation
Inverse Gaussian Population
Modified Signed Log Likelihood Method
Maximum Likelihood estimation
[1] Wise M.E., "Skew probability curves with negative powers of time and related to random walks in series", Statistica Neerlandica, 25(1971) 159–180.

[2] Wise M.E., "Skew Distributions in Biomedicine Including Some with Negative Powers of Time in Statistical Distributions in Scientific Work", vol. 2, D. Reidel, Dordrecht, Holland, (1975).

[3] Wise M.E., Osborn S.B., Anderson, J., Tomlinson R.W.S., "A stochastic model for turnover of radio-calcium based on the observed laws", Math. Biosci., 2(1968) 199–224.

[4] Lancaster T., "stochastic model for the duration of a strike", J. Roy. Stat. Soc. A, 135(1972) 257–271.

[5] Whitmore G.A., "The inverse Gaussian distribution as a model of hospital stay", Health Serv. R., 10(1975) 297–302.

[6] Schrödinger E., "Zür theorie der fall- und steigversuche an teilchen rnit Brownchar bewegung", Physikalische Zeitschrijt, 16(1915) 289-295.

[7] Tweedie M.C.K., "Statistical properties of inverse Gaussian distributions I", Ann. Math. Stat., 28(1957) 362-377.

[8] Tweedie M.C.K., "Statistical properties of inverse Gaussian distributions II". Ann. Math. Stat., 28(1957) 696-705.

[9] Chhikara R.S., Folks J.L., "The Inverse Gaussian distribution", Marcel Dekker, New York (1989).

[10] Seshadri V., "The Inverse Gaussian distribution: Statistical Theory and Applications", Springer, New York (1998).

[11] Searles D.T., "The utilization of a known coefficient of variation in the estimation procedure", J. Amer. Statist. Assoc., 59(1964) 1225–1226.

[12] Srivastava V.K., "On the use of coefficient of variation in estimating normal mean", J. Indian Soc. Agricultural Statist., 26(1974) 33–36.

[13] Chaubey Y.P., Sen D., "Estimator of mean in an inverse Gaussian population based on the coefficient of variation", J. Statist. Res., 42(2008) 1–16.

[14] Reid N., "Neyman from Life", Springer, New York (1982).

[15] Chhikara R.S., Folks J.L., "The inverse Gaussian distribution as a lifetime model", Technometrics, 19(1977) 461-468.

[16] Chhikara R.S., "Statistical Inference Related to the Inverse Gaussian distribution", Ph.D. dissertation, The Oklahoma State University (1972).

[17] Hsieh H.K., "Inferences on the coefficient of variation of an inverse Gaussian distribution", Communications in Statistics – Theory and Methods, 19(1990) 1589-1605.

[18] Chaubey Y.P., Sen D., Saha K.K., "On testing the coefficient of variation in an inverse Gaussian population", Statistics and Probability Letters, 90(2014) 121–128.

[19] Reid N., "Likelihood and higher-order approximations to tail areas: a review and annotated bibliography", Can. J. Statist., 24(1996) 141–166.

[20] Wu J., Wong A.C.M., Jiang G., "Likelihood-based confidence interval for a log-normal mean", Stat. Med., 22(2003) 1849–1860.

[21] Wu J., Wong A.C.M., "Improved interval estimation for the two-parameter Birnbaum-Saunders distribution", Compute. Stat. Data Anal., 47(2004) 809–821.

[22] Tian L., Wilding G.E., "Confidence interval of the ratio of means of two independent inverse Gaussian distributions", J. Stat. Planning Inference, 133(2005) 381–386.

[23] Sun Y., Wong A.C.M., "Interval estimation for the normal correlation coefficient", stat. Prob. Letters, 77(2007) 1652–1661.

[24] Cox D.R., Hinkley D.V., "Theoretical Statistics", Chapman and Hall, London (1979).

[25] Pierce D.A., Peters D., "Practical use of higher order asymptotic for multi-parameter exponential families", Journal of the Royal Statistical Society, Series B (Methodological), 54(1992) 701-737.

[26] Barndorff-Nielsen O.E., "Inference on full or partial parameters based on the standardized signed log likelihood ratio", Biometrika, 73(1986) 307-322.

[27] Barndorff-Nielsen O.E., "Modified signed log likelihood ratio", Biometrika, 78(1991) 557-563.

[28] DiCiccio T.J., Martin M.A., Stern S.E., "Simple and accurate one-sided inference from signed roots of likelihood ratios", Canadian Journal of Statistics, 29(2001) 67-76.

[29] Fraser D.A.S., Reid N., Wu J., "A simple general formula for tail probabilities for frequentist and Bayesian inference", Biometrika, 86(1999) 249–264.

[30] Ferguson T.S., "A Course in Large Sample Theory", Chapman & Hall, New York (1996).

[31] Tian L., "Testing equality of inverse Gaussian means under heterogeneity, based on generalized test variable", Comp. Stat. Data Anal., 51(2006) 1156–1162.