|
¦³¥ô¦ó°ÝÃDÅwªï¨Ó«H¼x¸ß
*
´£¨Ñ½u¤W¨ê¥dªA°È
½Ð¹w¥ý§iª¾!ÁÂÁÂ!

¡@ |
¡@
³Ì·sª©
Comprehensive Meta-Analysis Version
2
¡@
¬ÛÃö³nÅé
Power
And Precision (Power Analysis) Version 2
¡@
¡@
Comprehensive
Meta-Analysis 10¤j¥\¯à:
1.Work with a spreadsheet interface
2.Compute the treatment effect (or effect size)
automatically
3.Perform the meta-analysis quickly and accurately
4.Create a high-resolution forest plot with a single click
5.Perform a cumulative meta-analysis
6.Perform a sensitivity analysis
7.Assess the impact of moderator variables
8.Work with multiple subgroups or outcomes within studies
9.Assess the potential impact of publication bias
10.Work with subsets of the data
¡@
* for the Researcher
CMA is incredibly easy to learn and use, with a clear and
intuitive interface. The interactive guide will walk you
through all steps in the analysis, allowing new users to be
productive within minutes.
* for the Statistician
CMA was developed in collaboration with many of the
recognized experts in the field of meta-analysis, both in
the US and the UK. It includes a wide array of sophisticated
options for data entry, analysis, and display.
* for the Academic instructor
With CMA, the logic of meta-analysis comes alive. Use the
program to help explain complex issues, such as the impact
of study weights on the combined effect, the implications of
heterogeneity, or the distinction between fixed effect and
random effects models.
* Developed by a team of experts
CMA was developed by recognized experts in the field of
meta-analysis, both in the US and the UK.
¡@
¡@¡@Meta-analysis¤èªkªñ¦~°ê¤º¤w³vº¥¶}®iÀ³¥Î,¥»¤å´N¨ä°ò¥»·§©À¡BÀ³¥Î½d³ò¡B¾Þ§@¨BÆJ¤Î±`¥Îªº²Îp¾Ç¤èªk§@¤@ºîz¡C
1¡@Meta-analysis°ò¥»·§©À
Meta-analysis¤èªkªº«ä·Q¥i°l·¹¨ì20¥@¬ö30¦~¥N,³ÌªìÀ³¥Î©ó±Ð¨|¾Ç¡B¤ß²z¾Çµ¥ªÀ·|¬ì¾Ç»â°ì¬O¦b60¦~¥N,70¦~¥NªìLigh©MSmith´£¥X¤F¥i¥H¥Ñ¤£¦P¬ã¨sµ²ªG¶×Á`ì©l¸ê®Æ¶i¦æºî¦X¤ÀªR,1976¦~¥ÑGlassº¦¸©R¦WMeta-analysis¡C
Meta-analysis¤@µüªº·N«ä¬Omore comprehensive,§Y§ó¥[¥þ±ªººî¦X©Î¶W±`³W¡C°ê¤ºÄ¶¬°¤¸¤ÀªR©ÎîPµÑ¤ÀªR¡C³o¨Ç¦WºÙ³£ªí¥Ü³o¼Ë1Ó·§©À:§Y¹ï¥H©¹ªº¬ã¨sµ²ªG¶i¦æ¨t²Îªº©w¶q¤ÀªR¡CGlass¹ïMeta-analysis´£¥Xªº©w¸q¬O:¥Hºî¦X¤w¦³ªºµo²{¬°¥Øªº,¹ï³æÓ¬ã¨sµ²ªGªº¶°¦Xªº²Îp¾Ç¤ÀªR¤èªk¡CSacksµ¥´£¥Xªº©w¸q¬O:¹ï¥H©¹ªº¬ã¨sµ²ªG¶i¦æ²Îp¾Çªº¦X¨Ö©MÄYÂÔªººîz¤èªk¡Cù´ðµ¥»{¬°Meta-analysis¤èªk¬O¨Ì¾a·j¶°¤w¦³©Î¥¼µoªíªº¨ã¦³¬Y¤@¥i¤ñ¯S©Êªº¤åÄm,À³¥Î¯S©wªº³]p©M²Îp¾Ç¤èªk¶i¦æ¤ÀªR»Pºî¦Xµû»ù,¨Ï¦³¥i¯à¹ï¨ã¦³¤£¦P³]p¤èªk¤Î¤£¦P¯f¨Ò¼Æªº¬ã¨sµ²ªG¶i¦æºî¦X¤ñ¸û¡C
Meta-analysis¤èªkÀ³¥Î¨ìÂå¾Ç¦³Ãöªº¤å³¹²Ä1½gµoªí©ó1955¦~¡C§@ªÌºî¦X¤F15¥÷³æ¿W¬ã¨sµ²ªG,¹ï1
000¾l¦W¤£¦P¯e¯f±wªÌªA¥Î¦w¼¢¾¯ªºÀø®Ä¶i¦æ¤ÀªR,±o¥X¤F¦w¼¢¾¯¨ã¦³35
%Àø®Äªºµ²½×¡Cªñ¤Q¦h¦~¨Ó,Meta-analysis¦bÂå¾Ç»â°ìÀ³¥Î½d³ò¤é¯q¼sªx,¦b¶EÂ_¡BªvÀø¡B¦MÀI«×µû»ù¡B¤z¹w±¹¬I¡B¹w¨¾¹ïµ¦µ¥¤è±°_µÛ¿W¯Sªº§@¥Î,¬Æ¦Ü¦³¤H´£¥X¦b¶i¦æ¤åÄmºîz®É,À³¾¨¶q±Ä¥ÎMeta-analysis¤èªk¥H¨ú¥N¶Ç²Îªº´yz©Êªº¤èªk,±q¦Ó¥i¥H±o¨ì1ӳ̫᪺¥i¥H«H¿àªºµ²½×,¦]¦¹,¦³¤HºÙ¤§¬°¤ÀªRªº¤ÀªR©Î¥s¡§µ²ªG¬y¦æ¯f¾Ç¡¨,¥ç¦³¤HºÙMeta-analysis¬O§ó°ª1¯ÅÅÞ¿è§Î¦¡ªº¬ã¨s¡C
2¡@Meta-analysis¥Øªº¤Î¾A¥Î©Ê
¡@
¡@¡@Meta-analysis¤èªk¥Dn¸Ñ¨M¥H¤Uªº°ÝÃD:
1)¡@¼W¥[²Îp¥\®Ä:
´£°ª¹ïªì¨Bµ²½×ªº½×ÃÒ±j«×¤ÎÁ{§É©Ò¨£®ÄÀ³ªº¤ÀªRµû¦ô¤O«×¡C¥Ñ©ó³æÓÁ{§É¸ÕÅ穹©¹¼Ë¥»¤Ó¤p,Ãø¥H©ú½TªÖ©w¬YºØ®ÄÀ³,¦Ó³o¨Ç®ÄÀ³¹ïÁ{§É¨Ó»¡¤S¥i¯à¬O«nªº¡C¦pªGn¨D±q²Îp¾Ç¤W¨ÓªÖ©w©Î±Æ°£³o¨Ç®ÄÀ³,«h»Ýn¸û¤jªº¼Ë¥»,¦ÓY±Ä¥ÎMeta-analysis¤èªkn¤ñ1¶µ¤j³W¼Ò¡B¥N»ù°ª©ù¬Æ©Î¤£¤Á¹ê»Úªº¬ã¨s§ó¬°¥i¦æ,¦Ó¥B§â³\¦h¨ã¦³¥i¤ñ©Êªº³æÓÁ{§É¸ÕÅçµ²ªG¶i¦æ¦X¨Ö¤ÀªR,¥i¥H§ïµ½¹ï®ÄÀ³ªº¦ôpȩΧⴤ«×¡C
2)¡@¸Ñ¨M¦U¬ã¨sµ²ªGªº¤£¤@P©Ê:
¹ï¦P1Ó¬ã¨s°ÝÃD,¦UÓÁ{§É¸ÕÅçµ²ªG¥i¯à¤£ºÉ¤@P,¬Æ©Î¦s¦b¤Àª[ª§Ä³,§Q¥ÎMeta-analysis¤èªk¥i¥H±o¨ì¹ï¸Ó°ÝÃDªº¥þ±»{ÃÑ,§@¥X¬ì¾Çªºµ²½×¡C
3)¡@´M¨D·sªº°²»¡:
Meta-analysis¤èªk¥i¥H¦^µª³æÓÁ{§É¸ÕÅ礤©|¥¼´£¤Î©Î¬O¤£¯à¦^µªªº°ÝÃD,¤×¨ä¥Î©óÀH¾÷¹ï·Ó¸ÕÅç³]p©Ò±oªºµ²ªG¶i¦æºî¦Xµû»ù,¥i¥H´£¥X¤@¨Ç©|¥¼¬ã¨sªº·s°ÝÃD¡C
3¡@Meta-analysis¾Þ§@¨BÆJ¤Î¤èªk
3.1¡@¨î©w¸ÕÅç³]p®Ñ:
º¥ý´£¥XMeta-analysis¬ã¨sªº¥Øªº,§YÀÀ¸Ñ¨M¨º¨Ç¦s¦b¤Àª[ª§Ä³ªº«n°ÝÃD,©ú½T³æÓ¬ã¨sªº¤J¿ï©M±Æ°£¼Ð·Ç,¬d¾\¤åÄmªº¤èªk¤Î©Òn±Ä¨úªº²Îp¾Ç¤ÀªR¤èªkµ¥¡C
3.2¡@¬d§ä¤åÄm:
¥iɧU¹q¸£À˯Á,¤]À³¥J²Ó¬d§ä²{´Á´Á¥Z¥Ø¿ý¡Bºîz©Ê¤åÄm¥H¤ÎÁ{§É¸ÕÅçµn°O¸ê®Æ,¥²n®ÉÁٻݦV¸Ó¬ã¨s»â°ìªº±M®a¿Ô¸ß¡CÀ³¤O¨D¥þ±,¤£±o¿òº|¹ïµ²ªGµû©w¦³«n¼vÅTªº½×¤å¡C³v¤@¥J²Ó´_²ß¦³Ãöì©l¤åÄm,°O¿ý¬ã¨s¤u§@ªº¯S¼x,¥]¬A¤åÄmÃþ«¬¡Bµoªí¦~¤ë¡B³]p¤èªk¡B¼Ë¥»¤j¤pµ¥¡C
3.3¡@¬ÛÃö¬ã¨sªº«~½èµû©w©M¨ú±Ë:
¥i®Ú¾Ú¨ãÅ饨ªº©M±M·~ª¾Ãѵ¥¨î©w¥Xµû©w¼Ð·Ç,¦p½T©w¨ü¸Õª«¥óªº¼Ð·Ç¡B¼Ë¥»§t¶q³]p¡BÀH¾÷¤À²Õ¤èªk¡Bª¼ªkÆ[¹î¡B¨ü¸ÕªÌ¨Ì±q©Êµ¥¡Cç°£¤£º¡¨¬¼Ð·ÇªÌ,¥H½T«OMeta-analysisªº¦³®Ä©Ê¡C
3.4¡@¸ê®Æªººî¦X
3.4.1¡@½T©w¬ã¨s®ÄÀ³ªº´ú¶q«ü¼Ð¡C±N¦¬¶°¨ìªºp¶q¸ê®ÆÀËÅç²Îp¶qtÈ¡BuÈ¡BFÈ¡C¬ÛÃöÃö«Yªº´yz©Ê«ü¼Ð¬ÛÃö«Y¼Ær¤Îp¼Æ¸ê®Æªº²Îp«ü¼Ð:²v¡B¤ñȤñ(OR)¡B¬Û¹ï¦MÀI«×(RR)¡B£q2ȵ¥¶i¦æºî¦X¥[Åv,pºâ¥X¦X¨Ö«áªº¥§¡²Îp¶q¡CYì¬ã¨s¶¡´ú¶q«ü¼Ð¤£¤@P,«h»ÝÂà´«¬°²Î¤@«ü¼Ð¡C
3.4.2¡@¦P½è©ÊÀËÅç(homogeneity
test)¡C¹ï¦hÓ¿W¥ß¬ã¨sªº°ò½u¬Û¦ü¬Oºî¦X¤ÀªRªº«e´£,Yì¬ã¨s¶¡¨ã¦³¤@P©Ê,«K¥i¥H±N¦hÓ²Îp¶q¶i¦æ¥[Åv¦X¨Ö;Yì¬ã¨s¶¡¤£¤@P,ç°£¯S¤j¡B¯S¤pªº©Î¬O¤è¦V¬Û¤Ïªº²Îp¶q«á¦Aºî¦X¡C
±`¥Îªº¦P½è©ÊÀËÅç¤èªk¦³:
①
¹Ï¥Üªk
Pi=(Pci-Pti)/Pci
¦¡¤¤,Pi=i¶µ¬ã¨sªº¦³®Ä²v,Pci=i¶µ¬ã¨s¹ï·Ó²Õµo¥Í²v,Pti=i¶µ¬ã¨s¹êÅç²Õµo¥Í²v¡C
ø»s®y¼Ð¹Ï,Áa®y¼Ð»P¾î§¤¼Ð¤À§O¥Ü¹êÅç²Õ¹ï·Ó²Õªºµo¥Í²v¡CµøPi¤À§Gªº³W«ßÁͶÕ,¦pªG«D±`¤À´²ªí©úµL¦P½è©Ê¡C
②
²v®tQÀËÅç
¦¡¤¤,Yi=Pti-Pci,Wi=S-1i
Si=Pti(1-Pti)/nti+Pci(1-Pci)/nci, nti=i¶µ¬ã¨s¹êÅç²Õ¤H¼Æ,nci=i¶µ¬ã¨s¹ï·Ó²Õ¤H¼Æ,w=∑WiYi/∑Wi
QwȪñ¦üªA±qdf=K-1ªº£q2¤À§G¡C
③¤ñȤñ£q2¦P½è©ÊÀËÅç
£q2=∑Wi(1nORi)2-∑Wi(1nORi)2/∑Wi
¦¡¤¤,Wi=(1/ai+1/bi+1/ci+1/di),a,b¤À§O¬°¯f¨Ò²Õ»P¹ï·Ó²Õ¦³¼ÉÅS¥vªº¨Ò¼Æ,c,d¬°¨â²ÕµL¼ÉÅS¥vªº¨Ò¼Æ¡C
df=K-1=1
④
¤ñȤñQÀËÅç
Q=∑Wi(Yi-w)2
¦¡¤¤Yi=1n[Pti(1-Pci)/Pci(1-Pti)]
Wi=Si-1,Si=[ntiPti(1-Pti)]-1+[nciPci(1-Pci)]-1,nti,nci¬°i¶µ¯f¨Ò¹ï·Ó¬ã¨s¯f¨Ò²Õ»P¹ï·Ó²Õªº¨Ò¼Æ;Pti,Pci¬°i¶µ¬ã¨s¨â²Õªº¼ÉÅS²v¡C
QȪñ¦üªA±qdf=K-1£q2¤À§G¡C
⑤
¨âÅܼƬÛÃö«Y¼Æªº»ô©ÊÀËÅç
£q2=∑ni(1-r-2)2S2i,£o=K-1
⑥
¨â§¡¼Æ¤§®tªº»ô©ÊÀËÅç
£q2=KS2d/S2e,£o=K-1
⑦ ¨âÓ²v¤§®tªº»ô©ÊÀËÅç
(3)®ÄÀ³ºî¦X,¹ï¨ã¦³¤@P©Êªº²Îp¶q¶i¦æ¥[Åv¦X¨Ö,ºî¦X¦ôp¥X¥§¡²Îp¶q¡C¹ïºî¦X¦ôpªº²Îp¶q¶i¦æ²Îp®Ö¹ïÁ`©M²Îp§PÂ_¡C³Ì«ápºâ¬Y¨Ç²Îp«ü¼Ðªº95
%¥i«H°Ï¶¡¨Ã¥H¹Ï¥Üªí¥Ü¡C
3.5¡@ÆF±Ó«×¤ÀªR(sensitivity analysis):
¬°µû»ùMeta¤ÀªR¯u¹ê©Êªº«ü¼Ð¤§¤@,±Ó·P©Ê°ª»¡©úµ²½×¥¿½T©Ê±j,¥~¦b¯u¹ê©Ê¦n§YÀ³¥Î½d³ò¼s,¦³¸û°ªªº°Ñ¦Ò»ùÈ¡Cµû»ùÆF±Ó«×¥i±q¥H¤U´X¤è±¦Ò¼{:
① ì©l½×µÛªº«~½èµû»ù(quality assessment):Meta¤ÀªRªº±Ó·P©Ê¨ú¨M©ó¦UÓì©l½×µÛªº¬ì¾Ç©Ê©M¤èªkªºÄYÂÔµ{«×,°²©wì©l¸ê®Æ¸û®t,¨º»ò©Ò§Î¦¨ªºµ²½×«Ü¤Ö¦³¹ê¥Î©Ê¡Cì©l½×µÛªº«~½è¥Dn¥H¬ã¨s³]p¬O§_ÄY®æ²Å¦X©Ò³W©wªºn¨D¶i¦æ¿Å¶q;
② Åܧó¤èªk(varying method)¹ïµ²½×ªº¼vÅT¤j¤p;
③ ¿ï¾Ü©Ê°¾Êªº¦³µL¡C
·í¶i¦æ¤åÄmµûz®É,¤£¥i¯à§ä¥X¦³Ãö¬Y1°²³]©Ò¶i¦æªº¥þ³¡¬ã¨s,¦³¤HºÙ¤§¬°¡§ÀÉÂd°ÝÃD¡¨(file drawer
problem)[16],¦]¬°¨º¨Ç¹ï¬ã¨s°²³]µLÅãµÛ·N¸qªº¬ã¨sµ²ªG,©¹©¹³Q¬ã¨sªÌ¥á¦b¤@Ãä¡C¤]¦³¤H»{¬°,¤wµoªíªº¬ã¨s±`°¾¦V©ó¶§©Êµ²ªG,¦pªG1Ó¬ã¨s©úÅã¦a¹w¥Ü¤£¯à¥X²{¶§©Êµ²ªG®É,¸Ó¬ã¨s©¹©¹³Q©ñ±ó,¥BµLÅãµÛ·N¸qµ²ªGªº¬ã¨s³ø§i¤@¯ë¤£¥i¯à³Qµoªí,³o¥i¯à·|¼W¤jµo²{§ó¦h¶§©Êµ²ªGªº¢¹Ãþµoªí»~®t,¦]¦¹³\¦h¬ã¨sªÌ±`¹ï¤wµoªí©M¥¼µoªíªº¬ã¨sµ²ªG³æ¿W¶i¦æMeta-analysis¡C¤]¥ipºâ»Ý¦h¤Ö³±©Ê¬ã¨sµ²ªG¥i¯à¨Ïµ²½×°fÂà,§Y¥¢¦w¥þ¼Æ(fail-safe
N,Nfs),P¬°0.05©M0.01®ÉªºNfs¤À§O¬°:
Nfs0.05=(∑Z/1.645)2-S
Nfs0.01=(∑Z/2.33)2-S
∑Z¬°³æÓZȤ§©M,S¬°¥ÎStouffer'sªk¦X¨ÖÀËÅ窺¬ã¨sӼơC
3.6¡@³Ì«á§Î¦¨µ²½×,Á`µ²¦¨¤å
¤å¤¤À³¸Ô²Ó³¯z¤ÀªRªº¥Øªº;¤åÄm¬d§ä¤èªk¤Î¨ú±Ë¼Ð·Ç;©Òºî¦Xªº³æÓ¬ã¨sªº¯S¼x;»¡©ú©ÒÀ³¥Îªº²Îp¾Ç¤èªk;´£¨Ñ¥]§t¦³¦UÓ¬ã¨s²Îpµ²ªGªº¹Ïªí;´£¨ÑÆF±Ó«×¤ÀªRµ²ªG;µ²½×¥i¯à¹J¨ìªº°¾Ê¤Î³B²z¤èªk;°Q½×¤ÀªRµ²ªGÀ³¥Î»ùȵ¥¡C
4¡@Meta-analysisªº«~½èµû»ù
µû»ùMeta-analysis¤å³¹ªº¼Ð·Ç¥i¥H±q´XÓ¥Dn¤è±¦Ò¼{:
① ¬ã¨s³]p¬O§_¦X²z;
② ºî¦X¯à¤O¬O§_¥Î¤F«ê·íªº²Îp¾Ç¤èªk¶i¦æ¤ÀªR;
③ «ç¼Ë±±¨î°¾Ê;
④ ¬O§_°µ¤F¦P½è©ÊÀËÅç;
⑤ ¬O§_¦Ò¼{¨ì¤F±Ó·P©Ê¤ÀªR;
⑥ ¹ïµ²ªGªºÀ³¥Î¦p¦ó¡C
¦bÂå¾Ç»â°ìùØÀ³¥ÎMeta-analysis¤èªk,¥i¥H©w¶q¦aºî¦X¤ÀªR¦hÓ¬ã¨sµ²ªG,±o¥X¬ì¾Ç¡B¦X²z¡B¥i«Hªºµ²½×,¬°¶i¤@¨B¬ã¨s´£¨Ñ°Ñ¦Ò,¬°¯e¯fªº¹w¨¾¡BªvÀø¡B¶E Â_µ¥½Ñ¤è±´£¨Ñ§ó¥þ±§ó¥i¾aªº¨Ì¾Ú¡C
|