เอกสารการประชุมวิชาการและเสนอผลงานวิจัย มหาวิทยาลัยทักษิณ ครั้งที่ 19 2552 - page 158

2
„µ¦ž¦³»¤ª·µ„µ¦Â¨³Áœ°Ÿ¨Šµœª·‹´¥ ¤®µª·š¥µ¨´¥š´„¬·– ‡¦´ÊŠš¸É 19 ž¦³‹Îµže 2552
Abstract
The objective of this study is to compare the efficiency of two estimation methods. They are Jackknifing
method and Bootstrapping method. The parameter to be estimated are mean, variance, skewness and kurtosis of a
set of data. Both point estimation and interval estimation are calculated in this study. Data are generated using
Monte Carlo Simulation technique through R-program. In this study, data are simulated to have scale
contaminated normal distribution, exponential distribution and gamma distribution. For scale contaminated normal
distribution, the distributed parameter are variance( ), percent of contamination and scale factor and in this study
the data are generated specified to be 25, percent of contamination to be 10% and 30% and scale factor to be 5
and 10. For exponential distribution, the distributed parameter is and in this study the data are generated
specified to be 0.1, 0.5, 1 and 1.5. For gamma distribution, the distributed parameter are and , and in this study
the data are generated specified to be 3, 4, 6 and 8 and to be 0.1. For all specified value of sizes (n) are 100,
200, 300, 400, 500, 600, 700, 800, 900 and 1,000. The generated data were repeated 500 times under each
situations. The Mean Square Error (MSE) is a criterion to compare the point estimation and the confidence
coefficient is a criterion to compare the interval estimation. The results of this study can be summarized as follow:
Point Estimation Case :
1.When the data is scale contaminated normal distribution, Bootstrapping method has minimum MSE for
estimating mean, skewness and kurtosis. Jackknifing method has minimum MSE for estimating variance.
2.When the data is exponential distribution, Bootstrapping method has minimum MSE for estimating
skewness and kurtosis. Jackknifing method has minimum MSE for estimating mean and variance.
3.When the data is gamma distribution, Bootstrapping method has minimum MSE for estimating mean,
skewness and kurtosis. Jackknifing method has minimum MSE for estimating variance.
Interval Estimation Case :
It is found that every specified distributed parameter in every distribution, Bootstrapping method has
maximum confidence coefficient.
Keyword: Jackknife , Bootstrap , Point Estimation and Interval Estimation
‡Î µœÎµ
„µ¦«¹„¬µª·‹´¥—εÁœ·œŠµœ˜nµŠÇ Á¡ºÉ°®µ o°¦»žÁ„¸É¥ª„´ ž¦³µ„¦ ¤´„‹³Äo ¦³Á¸¥ª·›¸ šµŠ™·˜· š¸ÉÁ¦¸¥„ªn µ „µ¦
°œ»¤µœÁ·Š™· ˜· Ž¹É Šž¦³„°—oª¥ „µ¦ž¦³¤µ–‡nµ¡µ¦µ¤· Á˜°¦r ¨³„µ¦š—°¤¤˜·“µœ
„µ¦«¹„¬µ‡nµš¸É°„™¹Š¨´„¬–³ °Šž¦³µ„¦ Ž¹É ŠÁ¦¸¥„ªnµ‡nµ¡µ¦µ¤·Á˜°¦rœ´Ê œ ‹³š¦µÅ—o„Șn°Á¤ºÉ° ‹³˜o °ŠšÎµ
„µ¦Â‹Šœ´ ‹µ„š»„Ç ®œnª¥Äœž¦³µ„¦ Ĝ„¦–¸ š¸Éž¦³µ„¦¤¸œµ—Ä®n „µ¦Â‹Šœ´ ‹µ„š»„®œnª¥Äœž¦³µ„¦
‹³˜o°ŠÁ¸¥Áª¨µÂ¨³‡nµÄo‹nµ¥‹Îµœªœ¤µ„ —´Šœ´Êœ‹¹ŠÄo „µ¦Á¨º°„˜´ªÂšœÁ¡¸¥ŠµŠnªœ °Šž¦³µ„¦¤µ«¹ „¬µ ¨oªÄo
‡nµš¸Éŗo‹µ„˜´ ª°¥nµŠŽ¹ÉŠÁ¦¸¥„ªnµ ‡nµ™·˜· ޞ¦³¤µ–‡nµ¡µ¦µ¤·Á˜°¦r°Šž¦³µ„¦ Ž¹ÉŠ„µ¦ž¦³¤µ–‡nµ¡µ¦µ¤·Á˜°¦r‹³
µ¤µ¦™ÂnŠ°°„Å—oÁž}œ°Šž¦³Á£š ŗo„n „µ¦ž¦³¤µ–‹» — ¨³„µ¦ž¦³¤µ–nªŠ —´Šœ´Êœ‹¹Š‡ª¦Á¨º°„čo