2
µ¦¦³»
¤ª·
µµ¦Â¨³Á°¨µª·
´
¥ ¤®µª·
¥µ¨´
¥´
¬·
¦´Ê
¸É
19 ¦³Î
µe
2552
for each set of data under the condition assigned when the population was Exponential Distribution, Weibull
Distribution, Gamma Distribution and Beta Distribution that assigned different shape of distribution by skewness
and kurtosis. Sample sizes were 20, 30, 50 and 100 at the level of significance of the goodness of fit test of 0.05.
For this research, the data were simulated by the Monte Carlo method. This simulation was repeated until the
difference between of i+1 to i with least considered not different under the criteria of
0001 .0 ~ ~
1
d
i
i
X X
( i =
1,2,…,k , that i is the cycle of the experiment).
The result of this research was found that for skew-distributions for every level of skewness, the highest
percentage of success for transformation forms came from Folder Power and Guerrero-Johnson Transformations.
Keywords
: Transformation; Multinormality ; Weibull Distribution
Î
µÎ
µ
Ä{
»
´
¤¸
µ¦Î
µµ¦ª·
Á¦µ³®r
®¨µ¥´
ªÂ¦ (multivariate analysis) Á
o
µ¤µÄo
ĵª·
´
¥o
µn
µÇ ®¨µ¥
µ
µ ¹É
µ¦ª·
Á¦µ³®r
®¨µ¥´
ªÂ¦È
¤¸
®¨µ¥ª·
¸
µ¦Á¨º
°Äo
ª·
¸
µ¦ª·
Á¦µ³®r
®¨µ¥´
ªÂ¦
¹Ê
°¥¼n
´
¨´
¬³
°
o
°¤¼
¨
¸É
Î
µ¤µ«¹
¬µÂ¨³»
¦³r
°µ¦¡¥µ¦r
Ã¥µ¦ª·
Á¦µ³®r
®¨µ¥´
ªÂ¦ º
°ªn
µÁ}
ª·
¸
µ·
·
¸É
Äo
´
´É
ªÅ
Ã¥°µ³Äo
Ħ¼
µ¦ª·
Á¦µ³®r
ªµ¤Â¦¦ª®¨µ¥´
ªÂ¦ (multivariate analysis of variance) ¨³µ¦ª·
Á¦µ³®r
®´
¤¡´
r
°·
°¨ (canonical correlation analysis) Á}
o
Á¡ºÉ
°Ä®o
Åo
¤µ¹É
o
°¦»
¸É
¤¸
¦³·
·
£µ¡ ¤¸
ªµ¤¼
o
° ¨³ÁºÉ
°º
°Åo
o
°¤¼
¨³o
°Á}
ŵ¤ÁºÉ
°Å
n
µÇÄ®o
Á®¤µ³¤´
o
°¤¼
¨¸É
o
°µ¦«¹
¬µ ¹É
ĵ¦
ª·
Á¦µ³®r
´Ê
³o
°Î
µ¹
¹
缃
°Å
®¦º
°
o
°¤¤·
(assumptions) n
°µ¦´
·
ÄÎ
µµ¦ª·
Á¦µ³®r
o
°¤¼
¨¸É
¤¸
°¥¼n
¹É
µ¦ª·
Á¦µ³®r
µ¦³Á£¤¸
o
°¤¤·
ÁºÊ
°o
ªn
µ ´
ªÂ¦»
´
ª³o
°¤¸
µ¦Â·
®¨µ¥´
ªÂ¦ ®¦º
°´
ªÂ¦Ä
Ân
¨³Á³o
°Á}
°·
¦³n
°´
Á¡¦µ³®µ¦µªn
µ
o
°¤¼
¨¸É
Î
µ¤µ«¹
¬µ¤¸
µ¦Â¸É
Ťn
·
®¨µ¥´
ªÂ¦³Î
µ
Įo
¨µ¦ª·
Á¦µ³®r
¸É
Åo
Ťn
¼
o
°Â¨³
µªµ¤n
µÁºÉ
°º
°
Ã¥´É
ªÅ¨o
ª
o
°¤¼
¨¸É
Åo
°µ³Å¤n
Á}
ŵ¤
o
°¤¤·
n
µ ǸÉ
Î
µ®Åªo
缃
°µ
o
°¤¼
¨¦·
¸É
Åo
Ã¥n
ªÄ®n
³¤¸
Î
µªo
°¥Á¤ºÉ
°Á¸
¥´
o
°¤¼
¨¸É
Ťn
Åo
¤µµµ¦¨° ¨³®µ
o
°¤¼
¨Å¤n
Á}
ŵ¤
o
°¤¤·
´
¨n
µªÂ¨o
ª
°µÎ
µÄ®o
µ¦ª·
Á¦µ³®r
Ťn
¼
o
° (invalid) ¨³Î
揬n
µ¦¦»
¨µ¦ª·
Á¦µ³®r
¸É
·
¡¨µ ¨³
µªµ¤ÁºÉ
°º
° ¹É
µ¦Âo
{
®µÄ¦¸
¸É
o
°¤¼
¨¤¸
µ¦Â¸É
Ťn
·
®¨µ¥´
ªÂ¦º
° µ¦Äo
ª·
¸
µ¦Â¨
o
°¤¼
¨ (data
transformation) Ã¥µ¦Â¨
o
°¤¼
¨¸Ê
³Á}
µ¦¦´
n
µ
o
°¤¼
¨o
ª¥µ¦Äo
ª·
¸
µ¦µ·
«µ¦r
¸É
Ťn
´
o
° ¹
¤¸
´
·
·
®¨µ¥Çn
µÅo
Î
µÁ°ª·
¸
µ¦Â¨
o
°¤¼
¨Á¡ºÉ
°Âo
Å
{
®µ´
¨n
µª Á¡ºÉ
°Ä®o
o
°¤¼
¨¸É
Åo
¤¸
µ¦Â¨¼n
Á
o
µ¼n
µ¦Â·
®¨µ¥´
ªÂ¦¤µ¸É
»
Án
Äe
.«. 2003 Hans-Peter Piepho Åo
Î
µµ¦«¹
¬µÁ¸É
¥ª´
µ¦Â¨
o
°¤¼
¨Âµ¦ÂÂÂÁµ Ã¥
Äo
ª·
¸
µ¦Â¨
o
°¤¼
¨Â folded exponential (folded exponential transformation) µ¦Â¨
o
°¤¼
¨Â â¦r
¡µªÁª°¦r
(folded power transformation) µ¦Â¨
o
°¤¼
¨Â Guerrero-Johnson (Guerrero-Johnson
transformation) µ¦Â¨
o
°¤¼
¨Â Aranda-Ordaz (Aranda-Ordaz transformation) ¨³µ¦Â¨
o
°¤¼
¨Â