"ç¸é¢"ã®è©±ï¼ãã®ã¤ãã§ã«"21ä¸ç´ã®ç¸é¢ï¼MICï¼"ã®è©±ï¼ããããã¢åãï¼
ã©ãã§ããæ岳彦ã§ããæ¯åã®3DSã«ãã¼ãã£ã«ã³ã³ã½ã¼ã«ã®ãã½ãã¢ã³ã®éµããå¯ãã«å ¥ãã¾ããï¼ã¾ã ï¼é¢ï¼ã
ãã¦ã
ååã®è¨äºï¼
因果関係がないのに相関関係があらわれる4つのケースをまとめてみたよ(質問テンプレート付き) - Take a Risk:林岳彦の研究メモ
ã«ã¤ãã¾ãã¦ã¯æ²¢å±±ãã¯ãçãããã ã大å¤ãããã¨ããããã¾ãã*1ã大å¤æè¬ãã¦ããã¾ãã
ãã¦ãä¸è¨è¨äºã«ã¤ãã¦ãublftboãããããç¸é¢é¢ä¿ã®å®ç¾©ãæ¸ããã¦ããªãã®ã§ã¯ãï¼相関と因果 - Interdisciplinaryï¼ã¨ã®ãææãããã ãããã¾ããã
ãææã¯ç¢ºãã«ããã£ã¨ãã§ãã®ã§ãä»åã¯ãç¸é¢ãæ¦å¿µã«ã¤ãã¦ã¨ããã®ã¤ãã§ã«è¿å¹´ã«éçºããã"21ä¸ç´ã®ç¸é¢ï¼MICï¼"ã®è©±ã«ã¤ãã¦ç§ãªãã«æ¸ãã¦ã¿ããã¨æãã¾ãã
ï¼ä»¥ä¸ãããããã¢åãã®è©±ã«ãªãããããã¾ããããã¨ååã»ã©ã§ã¯ãªãã§ããããããªãã«é·ãã§ããï¼
åºç¾©ã®ãç¸é¢ãï¼ããé©åã«ã¯ãé¢é£ï¼associationãã¨å¼ã°ãããã®ï¼
ã¾ãã¯ãååã®è¨äºã®è£è¶³çãªã¨ãããã話ãã¯ããããã¨æãã¾ãã
ååã®è¨äºã«ããã¦ãç¸é¢ãã¨ããèªãç§ãã©ãããæå³ã§ç¨ãã¦ãããã¨ããã¨ããã¾ãæ·±ãã¯èãã¦ã¯ãã¾ããã§ããããæ¹ãã¦æåã«ããã¨ãå¾ããããã¼ã¿ã«ãããé ç®Aã¨é ç®Bã®éã«ä½ããã®é¢é£ãè¦ãããï¼ç¸é¢ããããã¨ããæå³ã§ç¨ãã¦ããã¨æãã¾ãã
ããã¯ããªããã£ããããç¨æ³ã§ãããªãåºç¾©ã®æå³ã§ã®ãç¸é¢ãã¨ããèªã®ä½¿ãæ¹ã¨è¨ãã¾ãï¼æ¥å¸¸è¨èªã«ããããç¸é¢ãã®èªã®ãã¥ã¢ã³ã¹ã®æ¹ãå¼·ãåæ ããç¨æ³ã¨ãè¨ããããããã¾ããï¼ã
Wikipediaの"correlation"の項ãè¦ã¦ã¿ãã¨ï¼
In loose usage, correlation can refer to any departure of two or more random variables from independence, but technically it refers to any of several more specialized types of relationship between mean values.
ã¨ããè¨è¿°ãããã¾ãããããã®ååã®"loose usage"ã«ããã"any departure of two or more random variables from independence"ã¨ããã¨ããããååã®è¨äºã«ããã¦ç§ã念é ã«ããã¦ãããç¸é¢ãã®èªã®æå³ããã¨ããã«ãªãã¾ãã
ããã§ã"departure of two or more random variables from independence"ã¨ããã¨ããã®æå³ãè¯ãåãããªãæ¹ãå¤ããããããªãã®ã§ããé¢é£ï¼associationãã¨ãç¬ç«ï¼independenceãã®é¢ä¿ã«ã¤ãã¦ã¡ãã£ã¨èª¬æãå ãã¦ã¿ã¾ãï¼
ãã¦ããããããå¾ããããã¼ã¿ã«ãããé ç®Aã¨é ç®Bã®éã«ä½ããã®"é¢é£"ãè¦ãããï¼è¦ãããªããã¨ããã®ã¯ãçµ±è¨å¦çã«ã©ã表ç¾ããããã§ããããï¼
ä¸è¬çã«ã¯ãAãèµ·ãã確çP(A)ã¨ãBãèµ·ãã確çP(B)ãããã³ãAã¨Bãåæã«èµ·ãã確çP(A,B)ã®é¢ä¿ãï¼
P(A,B) = P(A)P(B)
ã®ã¨ãã«ãAã¨Bã¯ã独立ï¼independenceãã¨å¼ã°ãã¾ããã¾ããAã¨Bãç¬ç«ã§ããã¨ãããããã®éã«ã¯ãï¼çµ±è¨çã«ï¼é¢é£ããªããã¨ããã¾ãã
ä¾ãã°ãï¼ã¤ã®ã³ã¤ã³A, Bãæããã¨ãã«ããããããªã¢ãã¨ãªã確çãããããã P(A=ãªã¢ã)=0.5ãP(B=ãªã¢ã) = 0.5 ãã§ããã¨ãã¾ãããã
ããã§ãAããªã¢ãã§Bããªã¢ããã¨ãªã確ç P(A=ãªã¢ã, B=ãªã¢ãï¼ããã P(A=ãªã¢ã, B=ãªã¢ãï¼= P(A=ãªã¢ã) x P(B=ãªã¢ã) = 0.5 x 0.5 = 0.25 ãã§ããã¨ãã«ã¯ããã³ã¤ã³Aãæãã¦ãªã¢ãã«ãªã確çãã¨ãã³ã¤ã³Bãæãã¦ãªã¢ãã«ãªã確çãã¯ãç¬ç«ãã§ãããé¢é£ããªããã¨ãããã¨ã«ãªãã¾ãã
éã«ã P(A=ãªã¢ã, B=ãªã¢ã) ã0.25 ããé¸è±ããå ´åã«ã¯ãã³ã¤ã³Aã¨ã³ã¤ã³Bã®ããã ã«ãä½ããã®é¢é£æ§ããæ¨æ¸¬ããããã¨ã«ãªãã¾ãã
ã¡ãªã¿ã«ä¸ã®å¼ãå¥ã®å½¢ã§æ¸ãã¨ï¼
P(A) = P(A|B)
ã¨ãæ¸ããã¨ãã§ãã¾ããããã¯ããBãã©ãã§ããããã¯Aã®ç¢ºçã«å½±é¿ãåã¼ããªããã¨ãããã¨ã示ãã¦ãã¾ããã¾ãå¥ã®è¨ãæ¹ãããã¨ããBã«é¢ããæ å ±ã¯ãAã«é¢ããäºæ¸¬ã®å½¹ã«ç«ããªãï¼ï¼ä½ã®æ å ±ããããããªãï¼ãã¨ããè¨ãæ¹ãã§ãã¾ãã
ã¤ã¾ããAã¨Bãç¬ç«ã§ãããã¨ããã®ã¯ãAã¨Bã®é¢ä¿ããé¢é£ããªãããå½±é¿ãåã¼ããªãããäºæ¸¬ã®å½¹ã«ç«ããªãããæ å ±ããããããªãããªã©ãªã©ã®æå³ã«å¯¾å¿ãããã¨ã«ãªãã¾ãã
ã¯ãã
ã¨ããããã§ããAã¨Bã®éã«åºç¾©ã®æå³ã§ç¸é¢ããããã¯æ°å¦çï¼çµ±è¨å¦çã«ããã¨ãAã¨Bã¯ç¬ç«ã§ãªããã¨ããç¶æ³ã«å¯¾å¿ãããã®ã«ãªãã¾ãã
ä¸è¬ã«ã¯ããã®æå³ã§ã®ãåºç¾©ã®ç¸é¢ãã«ã¤ãã¦ã¯ããç¸é¢correlationãããããé¢é£associationãã¨ããèªã使ãããã®ãæ®éã§ãã®ã§ãç´°ããè°è«ãããéã«ã¯åºå¥ãã¦ä½¿ããã¨ãæã¾ããã§ããããï¼ãã®æå³ã§ãååã®ç§ã®è¨äºã®ç¨èªæ³ã¯ãã¾ãè¯ããªãã¨è¨ãã¾ã*2ï¼
ã§ã¯æ¬¡ã¯ããåç¥ã®æ¹ãå¤ãã¨æãã¾ãããç義ï¼ã¨ãããããä¸è¬çãªç¨æ³ã«ãããï¼ãç¸é¢ãã«ã¤ãã¦èª¬æãã¦ã¿ããã¨æãã¾ãã
ãããããç¸é¢ãã¯ç´ç·çé¢ä¿ã®ææ¨ã§ãã
ä¸è¬ã«ãçµ±è¨å¦ã®æèã«ããã¦ãç¸é¢ãã¨è¨ã£ãå ´åã«ã¯ãããã¢ã½ã³ã®ç¸é¢ä¿æ°ãã«åºã¥ããã®ãæå³ãããã¨ãæ®ãã©ãã¨æããã¾ãã
Wikipediaの「相関係数」ã®é ãããã¢ã½ã³ã®ç¸é¢ä¿æ°ã®æ°å¦çå®ç¾©ãå¼ç¨ããã¨ï¼
ã¨ãªãã¾ãã
ãã¦ãã§ã¯ããã¢ã½ã³ã®ç¸é¢ä¿æ°ã®ç´æçãªæ§è³ªãè¦ã¦ã¿ã¾ãããã
ï¼ãã¢ã½ã³ã®ï¼ç¸é¢ä¿æ°ã®ç¹å¾´ã¯ããã¼ã¿éã®ç´ç·çé¢ä¿ã®ã¿ãè¦ã¦ãããã¨ã«ããã¾ãã
ç´ç·çãªé¢ä¿ãè¦ã¦ããã¨ãããã¨ã¯ãAã¨Bã®éã«ãæãããªé¢é£ï¼associationãããããããªå ´åã«ããç¸é¢ä¿æ°ï¼correlation coefficientï¼ã¯ä½ãå¤ã«ãªããããã¨ãããã¨ãæå³ãã¦ãã¾ãã
ãç¾èã¯ä¸è¦ã«ãããããªã®ã§ãWikipediaã®å³ãè¦ã¦ã¿ã¾ãããï¼
http://ja.wikipedia.org/wiki/%E3%83%95%E3%82%A1%E3%82%A4%E3%83%AB:Correlation_examples2.svg ããå¼ç¨
ããã§ããããã®å³ã¯ãã¼ã¿ã®æ£å¸å³ã表ãã¦ãã¦ãä¸ã®æ°åã¯ããããã®ç¸é¢ä¿æ°ã表ãã¦ãã¾ãã
ä¸ã®å³ããï¼
- ç´ç·çé¢ä¿ã§ã°ãã¤ããå ¨ããªãå ´åã¯ç¸é¢ä¿æ°ã¯1ï¼ã¾ãã¯-1ï¼ã«ãªã
- ç´ç·çé¢ä¿*3ã®å¾ãã®å¤§ããã¯ç¸é¢ä¿æ°ã®å¤§ããã«ã¯é¢ä¿ãªã
- ãã ãå¾ããå®å ¨ã«ãã©ããã®ã¨ãã¯ç¸é¢ä¿æ°ã¯ã¼ã*4
- ç´ç·çé¢ä¿ã«ããã¦ã°ãã¤ãã大ããã¨ç¸é¢ä¿æ°ã¯ãã®åå°ãããªã
- éç´ç·çãªé¢é£ãè©ä¾¡ãããå ´åã«ã¯ç¸é¢ä¿æ°ã¯ãã¾ãå½¹ã«ãããªã
ã¨ãããç¸é¢ä¿æ°ãã®ç¹å¾´ããããã¨æãã¾ãã
ã¯ãããããããç¸é¢ä¿æ°ãã¨ã¯ããããããã®ãªãã§ãã
ãã¦ãä¸è¨ã®ç¹å¾´ãç解ãã¦ä½¿ãã°ç¸é¢ä¿æ°ã¯å¤§å¤ä¾¿å©ãªãã®ã§ããããããä¸ã®ä¸ã®å ¨ã¦ããç´ç·çé¢ä¿ãã ã¨æããªããã¨è¨ããããããã¯ã¾ãããããã§ããã¾ãã
ããã§MICã§ããï¼ãã¤ï¼ã
21ä¸ç´ã®"ç¸é¢"ï¼MICã¨ã¯ï¼
MICï¼Maximum Information Coefficient : MIC@Wikipediaï¼ã¨ã¯ãããããã£ããè¨ãã¨ãã©ããªå½¢ã§ã対å¿å¯è½ãª"ç¸é¢"ä¿æ°ãã§ãã
2011å¹´ã®Scienceã§åºçãããReshef et al. 2011ã«ããã¦çºè¡¨ããããã®ã§ãそのときのScience誌の解説文ã§ã¯ãA correlation for the 21st centuryããªãã¦æ¸ãããããã¦ãã¾ãï¼ã¹ã´ã¤ã!ï¼ã
MICã¯ãã¢ã½ã³ã®ç¸é¢ä¿æ°ã®ããã«ã¯åç´ãªæ°å¼ã§ã¯è¡¨ãããã³ã³ãã¥ã¼ã¿ã«ãã£ã¦ã´ãªã´ãªã¨è¨ç®ããã¾ããåºæ¬çãªã¢ã«ã´ãªãºã ã¨ãã¦ã¯ããã¼ã¿æ£å¸å³ãæ§ã ãªæ°ã®ã°ãªããï¼ï¼è§£å度ï¼ã§åºåã£ã¦ãããªãããæ§ã ãªè§£å度ã®å¤ã«ããã¦相互情報量ãæ大ï¼ï¼åã°ãªããå ã«å«ã¾ãããã¼ã¿å¯åº¦ã®ã³ã³ãã©ã¹ããæ大ã¨ãªããããªã¤ã¡ã¼ã¸ï¼ã¨ãªããããªåºåãæ¹ã決å®ããããããè¦æ ¼åããã®ã¡ã®æ大ã®æ å ±éãMICã®å¤ã¨ãã¦é¸æãã¦ããããã§ã*5ã
ã¡ãªã¿ã«相互情報量ãæ°å¼ã§è¡¨ãã¨ï¼
ã¨ãªã£ã¦ãããç¬ç«ï¼ P(x,y)=P(x)p(y) ï¼ã®ã¨ãã«ã¯ç¸äºæ å ±éã¯ã¼ãã¨ãªããã¨ããããã¾ãã
MICã®ç¹å¾´ã¯ãã©ããªå½¢ã®associationã§ãå®éåã§ããã§ããã¨ããã«ããã¾ããä¾ãã°ãããªæãã§ãï¼
http://lectures.molgen.mpg.de/algsysbio12/MINEPresentation.pdf ããå¼ç¨ã»æ¹å¤
ããã¯ãä¸çªå·¦ã®åã®ã¿ã¤ãã®ãã¼ã¿ã®å ´åã«ãMICã¨ãã¢ã½ã³ã®ç¸é¢ä¿æ°ãããããã©ã®ãããªå¤ãã¨ããã示ãã¦ãããã®ã§ãããã¢ã½ã³ã®ç¸é¢ä¿æ°ã§ã¯æ¤åºã§ãã¦ããªããããªéç·å½¢ã®å ´åã«ããã¦ããMICã§ã¯é«ãå¤ã示ããã¨ããããã¾ããï¼ã¡ãªã¿ã«MICãã¨ãå¤ã®ç¯å²ã¯0ãã1ã¾ã§ã«ãªã£ã¦ãã¾ãï¼
MICã®ç¹å¾´ã¨ãã¦ããã¼ã¿ãã°ãã¤ãã«å¾ã£ã¦ãã®å¤ãä½ä¸ãããã¨ãæãããã¾ãï¼
http://lectures.molgen.mpg.de/algsysbio12/MINEPresentation.pdf ããå¼ç¨ã»æ¹å¤
ãã®è¾ºãã®æ§è³ªã¯ããã¢ã½ã³ã®ç¸é¢ä¿æ°ã®æ§è³ªãè¯ãåãç¶ãã§ãããç´æçã«ãéåæã®ãªããã®ã§ãã
MICã¯ããªãgeneralãªãã®ãªã®ã§ãåºæ¬çã«ã¯ã©ããªå¯¾è±¡ã«ã§ãé©å¿ã§ãã¾ãããã®ä¸ã§ãææãªå¿ç¨ä¾ã¨ãã¦ãéºä¼åçºç¾ã«ããããéç·å½¢çé¢é£ã®æ¤åºããæãããã¦ããããã§ãã
幸ããæè¿Rã§MICãç°¡åã«è¨ç®ããããã®"minerva"というパッケージãåºããããªã®ã§ãããã使ã£ã¦è¨ç®ã試ãã¦ã¿ããã¨æãã¾ãï¼
install.packages("minerva") library(minerva) data(Spellman) Spellman <- as.matrix(Spellman) res <- mine(Spellman,master=1,n.cores=1)
ããã§"Spellman"ã¨ãããã¼ã¿ã»ããã«ã¯ãCDC15 Yeast Geneã®4382åã®è»¢åç£ç©ã®éãæç³»åï¼23 time pointsï¼ã§è¨æ¸¬ãããã¼ã¿ãå ¥ã£ã¦ãã¾ãï¼è©³ããã¯こちらï¼ãmineé¢æ°ã«ããã¦MICã®å¤ãè¨ç®ããã¦ããã"res"ã«ã¯ãã®çµæãæ ¼ç´ããã¦ãã¾ãã
"res"ã®ä¸èº«ãè¦ã¦ãMICãé«ãå¤ã«ãªã£ã¦ãã2ã¤ã®è»¢åç£ç©ã®ä¾ãããã¯ã¢ãããã¦ã¿ãã¨ä»¥ä¸ã®ãããªãã¿ã¼ã³ã«ãªã£ã¦ãã¾ããï¼
ããããããã¢ã½ã³ã®ç¸é¢ä¿æ°ã§ã¯ä½ãå¤ã¨ãªãã±ã¼ã¹ã§ãããMICã§ã¯éç·å½¢çãªé¢é£ãæãããã¦ããããã§ãã
æ£ç´ã¡ãã£ã¨ãã»ãã¾ãããªãã¨æããªãã§ããªãã§ãããç 究ã®ãã¨ã£ãããããå¾ãåã«ã¯ååãªã®ããªãã¨ãæãã¾ãã
ã¾ã¨ã
ã¾ã¨ãã¾ãã
ä»åã¯ï¼
- ãï¼çµ±è¨çã«ï¼é¢é£ããªããã¨ã¯ãï¼çµ±è¨çã«ï¼ç¬ç«ã§ãããã¨ãããã¨
- ï¼çµ±è¨å¦ã®æèã§ï¼ãç¸é¢ãã¨ããã°ä¸è¬ã«ã¯ããã¢ã½ã³ã®ç¸é¢ï¼ä¿æ°ï¼ãã®ãã¨ãæã
- ï¼ãã¢ã½ã³ã®ï¼ç¸é¢ä¿æ°ã¯ç´ç·çé¢ä¿ããè©ä¾¡ãã¦ããªã
- ç´ç·çé¢ä¿ã«éããªããé¢é£åº¦ãã®ä¸è¬çãªææ¨ã¨ãã¦MICãªãã¦ã®ãããã¾ã
ã¦ãªè©±ã§ããã
次åã¯ããå æé¢ä¿ãããã«ããããããç¸é¢é¢ä¿ï¼ï¼ããæ£ç¢ºã«ã¯"çµ±è¨çé¢é£"ï¼ãçããªãã±ã¼ã¹ãã«ã¤ãã¦ã¾ã¨ãããã¨æãã¾ãã
é¢é£æ å ±ãªã©ã¾ã¨ã
- Correlation and dependence @Wikipediaï¼linkï¼
- 確çè«çç¬ç«æ§ @Wikipedia ï¼linkï¼
- MIC @Wikipedia (linkï¼
- MICã®å è«æï¼Reshef et al. 2011 in Scienceï¼ ï¼linkï¼
- åå·ã«è¼ã£ã¦ããMICã®è§£èª¬ï¼linkï¼
- Reshef et al. 2011ã®è£éºï¼link)
- MICã®è§£èª¬è³æï¼ãªã¹ã¹ã¡ï¼link
- Rã®MICã使ããããã±ã¼ã¸"minerva" ï¼linkï¼
ï¼é¢é£ããªãåãªã宣ä¼ï¼è¶ æå¦ããªãã«ããã¦ç§ãå¯ç¨¿ããã¦ããã ããæè¸èªãç·ã¨æ äºãè²·ãã¾ãï¼
ç§ãå¯ç¨¿ããã¦ããã ããã線と情事ããããã³ãã³è¶ ä¼è°2ã¨ä½µãã¦è¡ãããã超文学フリマãã«ããã¦ä»¥ä¸ã®è¦é ã«ããã¦è²©å£²ãããäºå®ã§ãã
ãè¶ æå¦ããªããin ãã³ãã³è¶ ä¼è°2:
2013å¹´04æ28æ¥ï¼æ¥ï¼10:00ã17:00å¹å¼µã¡ãã»ï¼ããã³ãã³è¶ ä¼è°2ãå ï¼ãã¼ã¹: ãã¤-03ã
ニコニコ学会βãªã©ã®ã¤ãã§ã«ã§ããç«ã¡å¯ãããã ããã°å¹¸ãã§ããä½åãããããé¡ããããã¾ãã
ï¼é販ã§ãè²·ãã¾ã
甘茶茂、しまおまほ、島田虎之介、関根美有ほか「線と情事 創刊号」 - タコシェオンラインショップ
創作・オリジナル 同人誌専門店 サッシ / 「線と情事」 / NECOfan
.
*1:å°å¿è ãªã®ã§ãã¯ãæ°ã伸ã³ããã¦ããã¬ãã·ã£ã¼ã§é¢¨éªãå¼ãã¾ãã
*2:ã¾ãããã£ãããç¸é¢ã¨å æãã¨ããå ´åã«ã¯åºç¾©ã®æå³ã§ã®ãç¸é¢ããå«æããã¦ããã¨èããã»ããè¯ãæ°ããã¾ãããããã«ãã¦ãã©ããã§ä¸åº¦å®ç¾©ãã¦ãããã»ããè¯ãã£ãã§ãã
*3:ç¸é¢ã®è©±ãã¡ã¤ã³ãªã®ã§ãããã§ã¯å帰ç´ç·ãæå³ãã¦ãã¾ãããå帰ã¨ç¸é¢ã®éãã«ã¤ãã¦ã¯âここãªã©
*4:ã¨æã£ããã©ãã°ãã¤ãããªãå ´åã«éã£ã¦ã¯å³å¯ã«è¨ãã¨ç¸é¢ä¿æ°ã®å¼ã®åæ¯ãã¼ãã«ãªãããè¨ç®ã§ããªãã£ã¦ã®ãæ£è§£ããï¼âã³ã¡ã³ãæ¬ã§ã®ãææã«å¾ãä¿®æ£
*5:ããããã¾ãèªä¿¡ãªãã§ãã詳ããã¯この解説資料ããã³原論文の補遺ãåç §æ¨å¥¨