357
O42
¨¦³
°µ¦ª·
Á¦µ³®r
µ·
·
£µ¥Äo
缃
°Å
µ¦Â·
Á¤ºÉ
°
o
°¤¼
¨¤¸
µ¦ÂÂÂè·
·
Effects of Statistical Analysis under Normal Distribution Assumption
on Data from Logistic Distribution
»
£µª¸
ª·
·
µ
1
¨³°»
£µ¡ ¤¼
¦r
ª´
¸
2
Supawadee Wichitchan
1
and Anupap Somboonsavatdee
2
´
¥n
°
µª·
´
¥¸Ê
¤¸
ª´
»
¦³r
Á¡ºÉ
°Á¦¸
¥Á¸
¥¦³·
·
£µ¡´
ª·
·
¸É
Äo
Î
µ®¦´
´
¦°µ¦ÂÂÂ
è·
·
µµ¦Â·
¨³«¹
¬µ¨¦³Äµ¦Äo
ª·
¸
µ¦¦³¤µn
µ´
¤¦³·
·Í
µ¦°¥Á¤ºÉ
°
ªµ¤¨µÁ¨ºÉ
°¤¸
µ¦ÂÂÂè·
·
¨µ¦ª·
´
¥¡ªn
µ ´
ª·
·
¸É
Äo
Î
µ®¦´
´
¦°µ¦ÂÂÂ
è·
·
µµ¦Â·
µ¤µ¦ª»
¤ªµ¤n
µ³Á}
°µ¦Á·
ªµ¤¨µÁ¨ºÉ
°¦³Á£¸É
1 Åo
Î
µ®¦´
Î
µ¨´
µ¦° ´
ª·
·
Shapiro Wilk Įo
n
µÎ
µ¨´
µ¦°¼
¸É
»
¦°¨¤µº
° ´
ª·
·
Anderson
Darling ´
ª·
·
Cramer Von Mises ´
ª·
·
Lilliefors ´
ª·
·
Chi-square ¨³´
ª·
·
Kolmogorov – Smirnov
µ¤¨Î
µ´
Ã¥n
µÎ
µ¨´
µ¦°
°»
´
ª·
·
³Á¡·É
¤¤µ
¹Ê
Á¤ºÉ
°
o
°¤¼
¨¤¸
µÄ®n
¹Ê
¨³Á¤ºÉ
°¦³´
´
¥Î
µ´
¼
¹Ê
[ Î
µ®¦´
µ¦¦³¤µn
µ´
¤¦³·
·Í
µ¦°¥Á¤ºÉ
°ªµ¤¨µÁ¨ºÉ
°¤¸
µ¦ÂÂÂè·
·
´Ê
Ä
¦¸
µ¦°¥Á·
¡®»
¨³¦¸
´
ªÂ¦°·
¦³¤¸
ªµ¤´
¤¡´
r
´
¹
¤o
ªn
µn
µ´
¤¦³·
·Í
µ¦°¥¸É
Åo
Ťn
Åo
¤¸
µ¦Â·
®¨µ¥´
ªÂ¦
Ân
µ¦¦³¤µn
µ´
¤¦³·
·Í
µ¦°¥ (£µ¥Äo
缃
°Å
ªn
µn
µªµ¤
¨µÁ¨ºÉ
°
H
¤¸
µ¦Â·
) Įo
¨°¨o
°´
n
µ¦·
Ä»
¦¸
]
Î
µÎ
µ´
: µ¦Â·
, µ¦ÂÂÂè·
·
, µ¦°¥Á·
Áo
Abstract
The purpose of this research is to compare the efficiency of test statistics used for screening logistic
distribution from normal distribution and study the effects on regression coefficient estimation when errors are
from logistic distribution. The results indicate that all test statistics can control the probability of type I error. With
respect to the power of test, Shapiro Wilk test gives the most power of test, followed by Anderson Darling,
1
·
·
¦·
µÃ £µª·
µ·
·
³¡µ·
¥«µ¦r
¨³µ¦´
¸
»
¯µ¨¦r
¤®µª·
¥µ¨´
¥ ¦»
Á¡ 10330
Master's Degree Student, Department of Statistic, Faculty of Commerce and Accountancy, Chulalongkorn University, Bangkok,
10330
2
°µµ¦¥r
¦. £µª·
µ·
·
³¡µ·
¥«µ¦r
¨³µ¦´
¸
»
¯µ¨¦r
¤®µª·
¥µ¨´
¥ ¦»
Á¡ 10330
Dr., Department of Statistic, Faculty of Commerce and Accountancy, Chulalongkorn University, Bangkok, 10330
Corresponding author : E-mail: