¥»ºô¶«Y¦C¥X¬ÛÃö±Ð¾Ç±Ð§÷¨Ñ°Ñ¦Ò
ªñ¥N²Îp¤èªk¤G:EM
Algorithm
EM Algorithm¬ODempester,
Laird, and Rubin©ó
1997¦~¦b
Journal of the Royal Statistics
Soeciety, Series B¤Wµoªíªº¤@Ó¿ìªk¡A¬O2ªñ¥N²Î
pµo®i¤W·¥«nªº¤@Ó¨½µ{¸O¡A¦Ó¦b°õ¦æ¤W±`»ÝɧU©ó²{¥N§Ö³tªº¹q¸£¡C
Á|¨Ò¨Ó»¡¡A·í¥Î¼Ë¥»¥§¡¨Ó¦ôºâ¥À¸sÅé´Á±æȮɡA¬°¹ï¦¹¤@¦ôºâªº»~®t¦³©ÒÁA¸Ñ¡A§ÚÌ
±`¥Î«H¿à°Ï¶¡(confidence
interval)ªº¿ìªk¨Ó°µ±À¦ô¡A¦¹®É±o¹ï¼Ë¥»¥§¡ªºsampling
distribution¦³©ÒÁA¸Ñ¡C¦b°ò¥»²Îp±Ð¥»¤W¡A·í¼Ë¥»©Ò¨Ó¦Ûªº¥À¸sÅé¡A¥i¥Î±`ºA¤À°t´yz®É¡A
¨äsampling
distribution¥i©Î¬°±`ºA¤À°t©Î¬°t¤À°t¡C¦ý·í¼Ë¥»©Ò¨Ó¦Ûªº¥À¸sÅé¡A¤£©y¥Î
±`ºA¤À°t´yz®É¡A§Ú̩Υιq¸£¼ÒÀÀ©Î¥Îº¥¶i¤ÀªRªº¿ìªk¥[¥H§JªA¡C·í¹ï¥À¸sÅ骺ÁA¸Ñ
¤£°÷²`®É¡Aº¥¶i¤ÀªRªº¿ìªk¬O¸û¦³®Äªº¤èªk¡A¬G¤¤¥¡·¥©w²z(Central
Limit Theorem)¡A
Edgeworth Expansion
(small sample theory)µ¥¿ìªk¤Î¨ä¥i¦æ©Ê¤Î¨îµ¥©ó¤åÄm¤¤¼s³Q
±´°Q¡A¤HÌÁö¤£¥þµM³ßÅw³o¨Ç¿ìªk¡A¦ý¤]§ä¤£¥X§ó²z©Êªº¤èªk¨Ó¨ú¥Nº¥¶i¤ÀªRªº¿ìªk¡C
¦Ó¦Û§Uªk½T¬O¤@Ó¬Û·í¨ã»¡ªA¤Oªº¤èªk¡A§ó´£¨Ñ¤F²Îp¤u§@ªÌ¥t¤@Ó´M§äsampling
distribution
ªº¿ìªk¡A¬G¦bªñ¦~¨Ó©ó¤åÄm¤¤¼s³Q±´°Q¡C
³q«UºtÁ¿: ÀH¾÷²{¶H
±Ð¾Ç¹q¸£¹êÅç:
ª½¤è¹Ï: ¦b¦¹¹êÅ礤¡AÂǥѿï¨úª½¤è¹Ï¤£¦Pªº°Ï¶¡ªø«×(bin width)¡A¨Ó±´°QCreated: January 1st, 1998
ª½¤è¹Ï¥i¯à·|±aµ¹§A¤£¦Pªº°T®§¡C
§A¤]¥i¤W¥t¤@Óºô¯¸¡AÂǥѴX²Õ¦³½ìªº¼Æ¾Ú¡A¨Ó¬Ý¥Î¤£¦P°Ï¶¡ªø«×
ªºª½¤è¹Ï±a©Ò±aµ¹§A¤£¦Pªº°T®§¡C
°²³]ÀË©w: ¦b¦¹¹êÅ礤¡AÂÇ¥ÑZÀË©w¨Ó±´°Q²Ä¤@«¬¦¡¿ù»~(Type I Error)¤Î
ÀË©w¤O(Power)
°jÂk: §Ú̳£ª¾¹Dpºâ¥§¡È®É¡A¦pªG¦³¤@Ó·¥¤j©Î·¥¤pªº¼ÆÈ¡A·í°£¥h
³oÓ¼ÆÈ«ápºâªº¥§¡È¡A©M¸ò¾Ú¥þÅé¼Æ¾Ú©Òpºâªº¥§¡È®t²§
·¥¤j¡C³o»¡©ú¤F§A®Ú¾Ú¥§¡È©Ò°µ¥Xªºµ²½×¡A¦³¥i¯à¬O¥u¨Ì¾Ú¤@Ó
¼ÆȪºµ²ªG¡C©Ò¥H±`«Øij¦b¹ê°È¤W§ÚÌÀ³¸Ó¦hºâ´XÓªí²{¤¤¥¡ÁͶժº
²Îp¶q¡A¦p¤¤¦ì¼Æµ¥¥H½T©w¬O§_¦³«ez¥i¯à©Êªºµo¥Í¡C¦Ó°jÂk¼Ò«¬
¬O¥§¡È¼Ò«¬ªº©µ¦ù¡A³oÓ±¡§Î§óÄY«¡A¤]¦³¥i¯à¬O¦]¬°¦ì¸m¦Ó²£¥Í
°ÝÃD¡C¦b¦¹¹êÅ礤¡A§A¥i¥[¤J¤@ӼƾڨÓÆ[¹î°jÂk½uºI¶Z¤Î±×²v¦p¦ó
¦]¤§§ïÅܪº¡C
µù:§A¥i¤W¥§¡È¤Î¤¤¦ì¼Æªººô¯¸¥hÅéÅç¤Wz°ÝÃD¡A§A¥i¥H¥Î·Æ¹«¦b
ª½¤è¹Ï¤W¥[¼Æ¾Ú©Î°Å¤Ö¼Æ¾Ú¨ÓÆ[¹î¥§¡È¤Î¤¤¦ì¼Æªº§ïÅÜ¡C
¦ÛÅܼƧt»~®t®Éªº°jÂk: ¦b°jÂk¤ÀªR¤¤¬O°²³]¦ÛÅܼƪº»~®t¬°¹s¡A¦¹®É¦]
À³ÅܼƤ¤©Ò§tªº»~®t»P¦ÛÅܼƪº¬ÛÃöªñ©ó¹s¬G§ÚÌ¥i¨Ï¥Î³Ì¤p¥¤èªk
¨Ó¨D±o°jÂk«Y¼Æ§_¡C¦b¦¹¹êÅ礤¡A§A¥i¨M©w¦ÛÅܼƪº»~®t¤j¤p¤Î
À³ÅܼƤ¤©Ò§tªº»~®t¤j¤p¡A¨ÓÆ[¹î°jÂk½uºI¶Z¤Î±×²v¦p¦ó¦]¤§§ïÅÜ
ªº¡C
¤G¶µ¤À°tªº¹Gªñ: ¦b°ò¦²Îp½Ò¡A§Ṳ́j³£·|¾Ç¨âÓ«nªº¾÷²v¤À¥¬¡A±`ºA
¤À¥¬¤Î¤G¶µ¤À¥¬¡C«eªÌ³Q¥Î¨Ó¼Ò«¬³sÄò«¬ªº¼Æ¾Ú¡A«áªÌ³Q¥Î¨Ó¼Ò«¬
Â÷´²«¬ªº¼Æ¾Ú¡C¦ý¬O·í§ÚÌnpºâ¤G¶µ¤À¥¬¤¤ªº¾÷²v®É¡A¦pªGnªº
¼Æȫܤj®É¡Apºâ¤W¬O¦³¬Û·íªº§xÃø¡A¦b°ò¦²Îp½Ò¤¤«Øijªº¸Ñ¨M
¿ìªk¡A¬O¥h¨Ï¥Î±`ºA¤À¥¬¤Î¤RªüªQ¤À¥¬¨Ó§@¹Gªñ¡C¦b¦¹¹êÅ礤¡A§A¥i
Åé·|±`ºA¤À¥¬¹Gªñªº¦nÃa¡C