ãã¼ããã¼ã¢ã©ã¤ã¢ã³ã¹é¨ 森ç°ã§ããææä¼å¡ã®ç²å¾æ½çããããã«é¢ãããµã¼ãã¹å åç·ã®æé©åãæ å½ãã¦ãã¾ãã
è¨äºã®å¯¾è±¡
- 仮説æ¤è¨¼ãéãã¦ä½ããæ¹åããããã¨æã£ã¦ãã人
- 仮説æ¤è¨¼ã®éã«ãã©ããããã®ãã¼ã¿ãéãããè¯ãããåãããªã人
ã¯ããã«
仮説æ¤è¨¼ã¨ã¯ã仮説ãç«ã¦ãããã証æããããã®ãã¼ã¿ãéããçå½ã確ããããã¨ãã§ããä»åã¯ä»®èª¬æ¤è¨¼ãè¡ãéã®æé ã¨ããã®æ¤è¨¼ã«å¿ è¦ãªãµã³ãã«ãµã¤ãºã®èãæ¹ã説æãã¾ãããµã³ãã«ãµã¤ãºã®è©±ã®ã¿é¢å¿ãããããã¯ãååãé£ã°ãããµã³ãã«ãµã¤ãºã®æ±ºãæ¹ããèªãã§ãã ããã
ç®æ¬¡
- è¨äºã®å¯¾è±¡
- ã¯ããã«
- ç®æ¬¡
- 仮説æ¤è¨¼ã®ã¤ãããã
- ãµã³ãã«ãµã¤ãºã®æ±ºãæ¹
- ã¾ã¨ã
- åèæ¸ç±
- 便å©ãµã¤ã
- 追è¨
仮説æ¤è¨¼ã®ã¤ãããã
1. 仮説ããã¦ã
仮説ããã¦ã¾ããç¥ããããã¨ãè¨ãåãå½¢ã«ããã°ä»®èª¬ã«ãªãã¾ãã
ä¾ãã°ãã¯ãã¯ãããã®å©ç¨è ãææä¼å¡ã«ãªããã£ããã¯æåã§ã¯ãªãç¾å³ããããªæçã®åçã§ãããããã¯ãã¯ãããå©ç¨è ã¯èªåã®é²è¦§ããã¬ã·ããå¾ããè¦è¿ãããããªã©ã§ãããã ããå¾è ã¯ä»®èª¬æ¤è¨¼ã®ããã®ä»®èª¬ã¨ãã¦ã¯å°ãä¸ååã§ãããã®çç±ã¯æ¬¡ã«èª¬æãã¾ãã
2. æ½ç/KPIãèãã
仮説ã¯æ ¹æ ã®ç¨åº¦ã¯ããã©çå½ã¯ä¸æã§ãããã®çå½ã確ãããããã«ä½ã®ãã¼ã¿ãéããå¿ è¦ããããèãã¾ãããããä¸çªã®é£æã§ãã
ãã¦ã¼ã¶ã¯èªåã®é²è¦§ããã¬ã·ããå¾ããè¦è¿ããããã¨ãã仮説ã®çå½ã¯ã©ã®ããã«ããã°åããã®ã§ãããããæ½çã¨ãã¦ã¯åç´ã«é²è¦§å±¥æ´ã®æä¾ãæãã¤ãã¾ãããããKPIã®è¨å®ã§è©°ã¾ãã¾ãã
ãã®ãããªã¨ãã¯ãç¥ããããã¨ãããããå°ãå ·ä½çã«ãã¾ãããã¨ãã°å²åã決ããããä½ãã¨æ¯è¼ããããã¾ããä¾ãã°ãå ¨ã¦ã¼ã¶ã®5%ãã¬ã·ããå¾ããè¦ç´ãããã¨æã£ã¦ãããããé²è¦§å±¥æ´ãå©ç¨ããã¦ã¼ã¶ã¯å©ç¨ããªãã¦ã¼ã¶ã«æ¯ã¹ã¦å©ç¨æéã20%伸ã³ãããªã©ã§ããæ©ãã å ´åã¯æ¬¡ã«è¿°ã¹ãã仮説æ¤è¨¼å¾ã®ã¢ã¯ã·ã§ã³ã決ãããã¨åããã¦èãããã¨ããããããã¾ãã
3. 仮説æ¤è¨¼å¾ã®ã¢ã¯ã·ã§ã³ã決ãã
å®éã«ãã¼ã¿ãéãããåãã«ããã¼ã¿ãéããå¾ã®ã¢ã¯ã·ã§ã³ã決ãã¾ãããã®ã¹ãããã«ããå ã®ãæ½ç/KPIããå¿ è¦ååãå¤æã§ãã¾ãããããããçµæãªãã°ãããããã¨ããã®ã決ããããªãããã§ããã°ããæ½ç/KPIãã¯ä¸ååãªã®ã§ããç´ãã¾ãã
ãµã¼ãã¹ã®æ¹åã®ããã«ãææ¨ã¨ãªãKPIã®æå³ã«ä¸»è¦³ãå«ã¾ããå ´åã¯ã¨ãã«æèãã¾ããäºå¾ã®è§£éã§ã¯ãµã³ã¯ã³ã¹ãã«ãããã¤ã¢ã¹ãããããå·éãªå¤æãé£ãããªãã¾ãããã¨ãã°ãé²è¦§å±¥æ´ã®å©ç¨è ã¯å ¨ä½ã®5%ãã¨ããçµæãã©ã®ãããªæå³ãæã¤ããæ¤è¨¼å¾ã«èããããã§ã¯ããã¾ããã
4. 対象ã決ãã
æ½çãKPIã決ã¾ãã¾ããã次ã«ãã®å¯¾è±¡è ã決ãã¾ãã
æçµçãªãµã¼ãã¹ã¸ã®å ¨ä½å½±é¿ãç¥ãããå ´åã¯ç´°ãããã¨ãèããå ¨ä½ãåå²ãã¦ABãã¹ããªã©ã§ãã¼ã¿ãéãã¾ãã
ä¸æ¹ã§ãä½ãããã®ä¾¡å¤ã®æç¡ã確èªãããå ´åãã¾ãã¯ããã®ã¦ã¼ã¶ã®è¡åãå¤ãããªãã®ã§ããã°ã価å¤èªä½ãç¡ãã®ã ãããã¨èããçæ³çãªã¦ã¼ã¶ãæ½çã®å¯¾è±¡ã¨ãããã¨ããããããã¾ãããªããªãã°å ¨ä½ã§ä¾¡å¤ã®æç¡ãæ¤è¨¼ããã¦ãã¾ãã¨å½±é¿ãå¸éããã¦è¨æ¸¬ãé£ãããªããã¨ãå¤ãããã§ãã
å ã®ã2. æ½ç/KPIãèãããã¨ããã®ã4.対象ã決ãããã§ã¯ãç¸é¢é¢ä¿ã¨å æé¢ä¿ã®åé¡ã«åå注æãã¦ãã ããããã¨ãã°å ã®ãé²è¦§å±¥æ´ãå©ç¨ããã¦ã¼ã¶ã¯å©ç¨ããªãã¦ã¼ã¶ã«æ¯ã¹ã¦å©ç¨æéã20%伸ã³ããã¨ãã仮説ã®æ¤è¨¼ã¯å®ã¯ãªããªãé£ããããããã®ãããªãã¼ã¿ãåããã¨ãã¦ããå©ç¨ããã¦ã¼ã¶ã¯ãã¨ãã¨å©ç¨æéãé·ãã¦ã¼ã¶ãã§ããå¯è½æ§ã¯é常ã«é«ãã§ãããã®åé¡ã«é¢ãã¦ã¯ä»åã®èª¬æã®ç¯çå¤ã®ãããã以ä¸è§¦ãã¾ããããèªæãªçãããªããããµã³ãã«ãµã¤ãºãè¨ç®ãããããåä»ã§ãã
5. ãµã³ãã«ãµã¤ãºãè¨ç®ãã
æ½çãKPIã対象ã決ã¾ãã¾ããããã¨ã¯å¿ è¦ãªãµã³ãã«ãµã¤ãºãè¨ç®ããã ãã§ããã¿ã¤ã ã¤ãºããã¼ãæéã¯æéã§è²´éã§ããPDCAãåãããã«ãè¡ãå½ããã°ã£ããã®è¨æ¸¬ã§ã¯ããã¾ããã
ãµã³ãã«ãµã¤ãºã®æ±ºãæ¹
ããããã¯ãµã³ãã«ãµã¤ãºã®æ±ºãæ¹ã説æãã¾ãã
ç§ãæ¥åã§è¡ã仮説æ¤è¨¼ã¯ã»ã¼å ¨ã¦ãABãã¹ãã¨ãã®å ã§è¿°ã¹ã仮説æ¤å®*1ã¨ããèãã®ãã¨ã§è¡ã£ã¦ãã¾ã*2ãããã¦ããã®éã®KPIã¨ãã¦ã¯ãå ¥ä¼ãã/ããªããããå©ç¨ããï¼ããªããã®ãããªãåç´ãª1,0ã好ã¿ã¾ãããã®ããããããã¯
- Aãå ¥ä¼ç10%ã®ç¾ã©ã³ãã£ã³ã°ãã¼ã¸ã
- Bãå ¥ä¼ç(以ä¸CVR)15%ãæå¾ ããæ°ã©ã³ãã£ã³ã°ãã¼ã¸(以ä¸LP)ã
ã¨ãã2ã¤ã®ãã¼ã¸ããBã¯AãããCVRã5%é«ãã ãã*3ãã¨ãã仮説ã®ãã¨ABãã¹ããããã®çµæãå¤æããã®ã«å¿ è¦ãªãµã³ãã«ãµã¤ãºnããä¿¡é ¼åºéã®èãããã¨ã«èª¬æãããã¨æãã¾ãã
æ¤è¨¼ã¨ãã¦ABãã¹ããããã¤ããããªãå ´åã§ãããã®å ã®èª¬æãç解ãããã¨ã§é©åãªãµã³ãã«ãµã¤ãºã決ããäºãã§ããããã«ãªãã¨æãã¾ãã
çããå ã«
ããããå ã®èª¬æã¯å°ãé·ããªãã¾ããå°ãã§ãç解ãããããããããããæ±ãããã¨ãããµã³ãã«ãµã¤ãºnã¨ã¯ä½ãããããã«æç« ã§è¡¨ç¾ãã¾ãã注æãã¹ãç¹ã¨ãã¦ãæ±ãããµã³ãã«ãµã¤ãºnã¯ãå·®ããããã¨ãã確èªããããã®ãµã³ãã«ãµã¤ãºã§ããããXã®å·®ããããã¨ãã確èªããããã®ãµã³ãã«ãµã¤ãºã§ã¯ããã¾ãã*4ã
確çp1ã¨ç¢ºçp2ã«X以ä¸ã®å·®ããããªãã°nåã®ãµã³ãã«ãµã¤ãºã§æ確ãªå·®ãã§ãã ãããããå·®ãåºãªãã®ã§ããã°X以ä¸ã®å·®ã¯ç¡ãã®ã ãã*5ã
ãµã³ãã«ãµã¤ãºã決ããäºã¤ã®è¦ç´
ä»åã®ABãã¹ãã«å¿ è¦ãªãµã³ãã«ãµã¤ãºã¯ãäºã¤ã®å¹³åå¤*6ãã¨ãå¿ è¦ãªç¢ºåº¦ãã§ãã¾ãã¾ãããã®ããã¾ãã¯ãäºã¤ã®å¹³åå¤ãã¨ãå¿ è¦ãªç¢ºåº¦ããããããä½ãæå³ãããã説æãã¾ããç¶ãã¦äºåç¥èã¨ãã¦ãäºé åå¸ãã®æ¦å¿µã説æãããå¾ããå¿ è¦ãªç¢ºåº¦ããå解ãã¦ç¾ãããαã¨Î²ãã説æããæå¾ã«ãµã³ãã«ãµã¤ãºãè¨ç®ãã¾ãã
ãäºã¤ã®å¹³åå¤ãã¨ã¯ä½ã
ä»åã§è¨ãã°10%ã¨15%ã§ãããµã³ãã«ãµã¤ãºã®è¨ç®ã«ã¯ãã®äºã¤ã®å¤ãå¿ è¦ã§ããçç±ã«ã¤ãã¦ã¯å°ãå ã®ãβãã§èª¬æãã¾ãã
ãå¿ è¦ãªç¢ºåº¦ãã¨ã¯ä½ã
確度ã¨ã¯ä¸è¨ã®äºã¤ã®ç¢ºããããã§ãã
(1) Aã¨Bã«å·®ããªã
(2) Aã¨Bã«å·®ããã
ãã®ä¸ã§ã(1)ã§ããã«ãããããããå·®ããããã¨ãããã¨ãã第1種ã®èª¤ããã¨ããããã®ç¢ºçããαãã¨ããã¾ã*7ã
ããã¦(2)ã§ããã«ãããããããå·®ããªããã¨ãããã¨ãã第2種ã®èª¤ããã¨ããããã®ç¢ºçããβãã¨ããã¾ããè¨ãæããã¨Î±ã¯å½é½æ§ã®ç¢ºçã§ãããβã¯å½é°æ§ã®ç¢ºçã§ãã
ã¤ã¾ããµã³ãã«ãµã¤ãºã決ããè¦ç´ ã®ä¸ã¤ã®ã確度ãã¨ã¯Î±ã¨Î²ã®äºã¤ã§ããããå¿ è¦ãªç¢ºåº¦ãã¨ã¯Î±ã¨Î²ãããããã©ã®ç¨åº¦ã¾ã§è¨±å®¹ããããæå³ãã¾ãã
ãAã¨Bã«å·®ããªããã¨ã¯ä½ã
å®ã®ã¨ãããAã¨Bã«å·®ããªããã¨ã¯ãBã¯Aã§ç¡ãã¨ã¯è¨ããªãããæå³ãã¾ããæ¶æ¥µçãã¤åããã©ã表ç¾ã§ãããããã«ã¯æå³ãããã¾ããå®éã®ã¨ããããããã¼ã¿ãéãã¦ããå®å ¨ã«å·®ããªããã¨è¨ãåãã¾ããããã®ããä½ãã®å·®ãå¤æããéã«ã¯ã¾ãããBã¯Aã¨çããã ãããã¨ããåæã«ç«ã¡ããã¼ã¿ãéãããBã®ãã¼ã¿ãè¦ãéãã確ççã«Aã¨çããå¯è½æ§ãååã«ãããç¶æ ãªãã°ãããããã£ã¦ãå·®ããªããã¨çµè«ãã¾ãããããããã¨ãBã®çµæãAã§ããã«ãã¦ã¯é¸è±ãã¦ããããã©ããã§å·®ã®æç¡ãçµè«ããããã§ããããã¦ãBã®ãã¼ã¿ãè¦ãéãã確ççã«Aã¨çããå¯è½æ§ãå°ãªããå ´åã«ã¯ãå·®ããããã¨çµè«ãã¾ããä½ãæã£ã¦ãé¸è±ãã¨ãããã¯ãã®ãã¨ã®ãαãã§èª¬æãã¾ãã
ãã®ãããªææ³ãã仮説æ¤å®ãã¨è¨ãã¾ããããããå ã«ç¶ã説æã¯ããã®ã仮説æ¤å®ãã¨ããèãã«åºã¥ãå·®ã®æç¡ãçµè«ãã¾ãããã詳ããç¥ãããæ¹ã¯ã仮説検定 - Wikipediaããèªãã§ãã ããã
äºé åå¸
ãã®å¾ã«ç¶ããαã¨Î²ãã®èª¬æã®ããã«ãäºåç¥èã¨ãã¦äºé åå¸ã¨ããæ¦å¿µã説æããã¾ãã
å è¦ãã説æãããã¨ãäºé åå¸ã¨ã¯
確çpã§æåããè¡çºãnåæ½è¡ããå ´åã®æåæ°ã®ç¢ºçåå¸
ã§ãã
ãã¨ãã°ãè£è¡¨ãã¨ãã«50%ã®ç¢ºçã§åºãã³ã¤ã³ã100åæããå ´åã表ãåºãæ°ã¯ã´ã£ãã50ã«ãªãã¨ã¯éãã¾ããããããåºãã»ããç¨ã§ããã§ã¯ä½åãä½%ã®ç¢ºçã§åºç¾ããã®ã§ãããããããã表ãããã®ãä¸ã®å³ã§ãããäºé åå¸ã§ãã
縦軸ã確çã横軸ã表ã®æ°ã«ãªããé¢ç©ãåè¨ããã¨1(=100%)ã«ãªãã¾ãããããã¿ãã¨å¹³åã50ã¨ãã¦ãã ãããã40â60ã®ç¯å²ã«åã¾ããã¨ããããã¾ãã
åãèãã§ããå ¥ä¼ãã/ããªãã ããããã1ã¨0ã§è¡¨ããå ¥ä¼ãã確çãpã¨ããå ´åããäºé åå¸ã§è¡¨ããã¨ãã§ãã¾ããä¸ã®å³ã¯200人(n)ãCVR10%(p)ã®LPã«æ¥è§¦ããããã«ãæçµçãªå ¥ä¼æ°ãã©ã®ããã«ã°ãã¤ããã示ãã¦ãã¾ããå®è£ ã®é½åä¸æ²ç·ã«ãªã£ã¦ãã¾ãããå®éã¯ä¸ã®å³ã®ããã«é段ç¶ã«åå¸ãã¦ãã¾ããã¾ããã¹ã©ã¤ãã移åãããã¨ã§nãå¤æ´ã§ãã¾ããnã¨åå¸ã®é¢ä¿ã確èªãã¦ã¿ã¦ãã ããããªãã赤ãå¡ãã¤ã¶ãããé¨åã®èª¬æã¯å¾ã»ã©ãããããæ°ã«ããå¿ è¦ã¯ããã¾ããã
αã¨Î²
äºåç¥èã®åå¼·ãçµããããããã¯Î±ã¨Î²ã«ã¤ãã¦èª¬æãã¾ãã復ç¿ã¨ãã¦ãããããã®æå³ã¯ãã®ãããªãã®ã§ããã
- αã¨ã¯ãçå®ã¨ãã¦å·®ããªãã«ãé¢ããããå·®ãããã¨çµè«ãããå½é½æ§ã®ç¢ºç
- βã¨ã¯ãçå®ã¨ãã¦å·®ãããã«ãé¢ããããå·®ããªãã¨çµè«ãããå½é°æ§ã®ç¢ºç
α(å½é½æ§ã®ç¢ºç)
αãå ¥ä¼æ°ã«å½ã¦ã¯ãã¦èãã¾ãã
200人ãCVR10%ã®ç¾LPã«æ¥è§¦ããå ´åã®å ¥ä¼æ°ã¯20人ãä¸å¿ã«ã°ãã¤ãã¾ããããã¦é¢ç©ã®(ã ããã)95%ã12人ãã28人ã®éã«ããã¾ããã¨ãäºé åå¸ã«ãã示ããã¦ãã¾ãããã®12â28ã®é åãã95%ä¿¡é ¼åºéãã¨å¼ã³ã¾ããããã¯äºã¤ç®ã®å³ã®èµ¤ã§å¡ãã¤ã¶ããã¦ããªãé¨åã«ç¸å½ãã¾ããããã¯ã¤ã¾ã赤ãé¨åã®é¢ç©ã¯5%ã¨ãããã¨ã«ãªãã¾ãã
Bã®å ¥ä¼æ°ãè¨æ¸¬ãããã®çµæããã®èµ¤ãé¨åã«è©²å½ããå ´åã¯ãå·®ããããã¨çµè«ããã¨æ±ºãããªãã°ãããã¯ãαã5%ã«è¨å®ãããäºãæå³ãã¾ãã
ãããBã®çå®ã®CVRãæå¾ ãã¦ããããª15%ã§ãªãAã¨åæ§ã®10%ã§ããå ´åãBã®è¨æ¸¬å¾ã®å ¥ä¼æ°ã¯5%ã®ç¢ºçã§èµ¤ãé¨åã«ãªãããããã®æã¯ãçå®ã¨ãã¦å·®ããªãã«ãé¢ããããå·®ãããã¨çµè«ããããã¨ã«ãªããå½é½æ§ã®ç¢ºç=α=5%ã¨ãªãã¾ãã
ãªãã決ããã¨è¡¨ç¾ããããã«ãαãä½%ãããã¯æ£æçã§ãã赤ãé¢ç©ãå°ãããªãããã«Î±ãè¨å®ããã°å½é½æ§ã®ç¢ºçã¯æ¸ãã¾ããããã®åãæ¬å½ã«å·®ãããå ´åããå·®ããªããã¨çµè«ãããããªãã¾ããä¸è¬çã«ã¯Î±ã1%ã5%ã10%ã¨ãããã¨ãå¤ãã§ãã
β(å½é°æ§ã®ç¢ºç)
ç¶ãã¦Î²ã®èª¬æããã¾ã
αã¯ä¸ã¤ã®åå¸ã§å®çµãã話ã§ãããã¨ããã®ã¯Î±ã¯æ¯è¼ãã¦èããåé¡ã§ã¯ãªãããã§ããããã«å¯¾ãã¦Î²ã¯æ¬å½ã¯åå¨ããå·®ãè¦éãã¦ãã¾ã確çã§ãããäºã¤ã®åå¸ãå¿ è¦ã§ããä»åã§ããã°CVR10%(A)ã¨CVR15%(B)ã®äºã¤ã§ãã
ã¾ãå ç¨ã®å³ã«ãCVR15%ã®åå¸ããªã¬ã³ã¸ã®ç·ã§å ãã¾ããã¾ããå ¥ä¼æ°ãå ¥ä¼çã«ç½®ãæããåå¸ãåããã¦ç¤ºãã¾ã*8ãç°è²ã®é¨åã«ã¤ãã¦ã¯ããã«èª¬æããã®ã§æ°ã«ããªãã§ãã ãããã¨ã¯ããå¯ãã®è¯ãæ¹ã®ããã«è¨ãã¾ãã¨ããã®ç°è²ã®é¨åãβã§ãã
復ç¿ã«ãªãã¾ããβã¨ã¯ãçå®ã¨ãã¦å·®ãããã«ãããããããå·®ããªãã¨çµè«ãããå½é°æ§ã®ç¢ºçã§ããä»åã®ä¾ã«ãã¦ã¯ããã¨ãBã®çå®ã®CVRã¯15%ã«ãé¢ããããn人ãæ¥è§¦ããçµæã®å ¥ä¼æ°ãã¿ã¦ãAã¨å·®ããªãã¨èª¤ã£ã¦çµè«ããã確çã§ãããã®æã«ãä½ããã£ã¦ãAã¨å·®ããªããã¨ãããã¯å ã«è¿°ã¹ãαã§æ±ºã¾ãã¾ããã¤ã¾ãβã¯Î±ã«ãã£ã¦å¤ããã¾ã*9ã
å ¥ä¼ä»¶æ°åå¸ã®ç°è²ã®é¨åã¯ãCVR10%æã«95%ã®ç¢ºçã§çºçããã人æ°ã®ãCVR15%æã®çºçå²åãã示ãã¦ãã¾ãããã®å²å(=確ç)ã¯å³ã®å±±å ¨ä½ã100%ã¨ããæã«39%ã¨ãªãããã®39%ãβã§ããβã39%ã¨ã¯ã¤ã¾ããBã®çå®ã®CVRã15%ã®å ´åã200人ã®ãµã³ãã«ãµã¤ãºã§ã¯39%ã®ç¢ºçã§ãBã¯A(10%)ã¨å·®ããªããã¨çµè«ãããäºãæå³ãã¾ã*10ã
ããããªããæ¬å½ã«å·®ãããã«ãé¢ãããã39%ã®ç¢ºçã§å·®ããªãã¨ãã¦ãã¾ãã®ã§ããã°ã(2) Aã¨Bã«å·®ãããããååãªç¢ºåº¦ã§çµè«åºæ¥ã¦ããã¨ã¯è¨ãã¾ãããã§ã¯ãã®Î²ã5%ã«ã¾ã§æ¸ããã«ã¯ã©ã®ããã«ãããããã§ãããããããã¾ã 試ãã¦ããªããã°æå¾ã®å³ã®ã¹ã©ã¤ããåããã¦n(ãµã³ãã«ãµã¤ãº)ã®å¤ãå¢ããã¦ã¿ã¦ãã ãããããã¨nã®å¢å ã«ä¼´ããå ¥ä¼ç¢ºçåå¸ã®å±±ãéè§ã«ãªããçµæã¨ãã¦ç°è²ã®é¢ç©ãæ¸ã£ã¦ãããã¨ããããã¾ããããã¯ã¤ã¾ãβãå°ããããã«ã¯nãå¢ããã°è¯ãã¨ãããã¨ã§ã*11ã
ã¾ããå³ãè¦ãã¨ãããããã«ãβã¯äºã¤ã®åå¸ã®å±±ã®è·é¢(å¹³åå¤ã®å·®)ã«ããå¤åãã¾ãããããB(ãªã¬ã³ã¸ã®ç·)ã15%ã§ãªã30%ã§ããã°å±±ã®è·é¢ã¯ãã£ã¨é¢ããããã«ã¨ããªãβã®å¤ãå¤åãã¾ãããã®ãã¨ãããé©åãªãµã³ãã«ãµã¤ãºããæ±ããã«ã¯ãäºã¤ã®å¹³åå¤ããäºåã«æ±ºããå¿ è¦ããããã¨ããããã¾ãã
ããã¾ã§ã®è©±ãæ´çããã¨ãã®ããã«ãªãã¾ãã
- α = æ£æçãªç¢ºç
- β = f(α, äºã¤ã®å¹³åå¤, ãµã³ãã«ãµã¤ãº)
ãã®ãã¨ããããé©åãªÎ²ã«ããã«ã¯ããµã³ãã«ãµã¤ãºãå¤åãããã°è¯ããã¨ããããã¾ãã
ãµã³ãã«ãµã¤ãºã®è¨ç®
ä»ã¾ã§ã®è©±ãå ã«ãαã¨Î²ãã¨ãã«5%ã¨ãªããããªnãè¨ç®ãã¾ããèãæ¹ã¨ãã¦ã¯ãAã®å³å´2.5%ã¨Bã®å·¦å´5%ã®æ°å¤ãä¸è´ãããããªnãæ±ãã¾ããããããã°Bã®å¹æã15%(以ä¸)ã§ããå ´åã«ååãªç¢ºçã§å·®ãæ¤ç¥ã§ããã ãã®ãµã³ãã«ãµã¤ãºãæ±ãããã¨ãã§ãã¾ã*12ã
è¨ç®ã«ã¯äºé åå¸ã®æ£è¦åå¸ã¸ã®è¿ä¼¼ãå©ç¨ãã¦ãã¾ã*13ãè¨ç®ã®ããã®Pythonã³ã¼ãã¯ä»¥ä¸ã®ããã«ãªãã¾ãã
import sympy as sp p = 0.1 diff = 0.5 n_s, p_s, diff_s = sp.symbols("n_s p_s diff_s") a_mean = n_s * p_s a_var = (a_mean * (1 - p_s)) a_sd = sp.sqrt(a_var) b_p = p_ + p_s * diff b_mean = n_s * b_p b_var = (b_mean*(1 - b_p)) b_sd = sp.sqrt(b_var) a_right = (a_mean + 1.96 * a_sd) b_left = (b_mean - 1.65 * b_sd) param = [(p_s, p), (diff_s, diff)] print(sp.solve((sp.Eq(a_right.subs(param), b_left.subs(param))), n_s)[1]) => 554.289654454896
ããã«ããã555人ã¨ããçµæãå¾ããã¾ããããã®äººæ°ãå ç¨ã®å³ã«è¨å®ããããå¡©æ¢ ã§ãããã¨ã確èªãã¦ã¿ã¦ãã ããã
ã¾ã¨ã
ååã¯ä»®èª¬æ¤è¨¼ã®ã¤ããããããå¾åã¯ãã®éã«å¿ è¦ãªãµã³ãã«ãµã¤ãºã®èãããã説æãã¾ããããä½ã®ããã«ä½ãè¨æ¸¬ããããã¯æã大åãªããã¤é£ããåé¡ã§ããããããããããã決ãããã¨ãåºæ¥ããªãã°ããã®æã¯ååãªãµã³ãã«ãµã¤ãºãä¼´ãæ£ããå¤æããªããã°ããã¾ãããã¾ãéä¸ã§è¿°ã¹ãã¨ãããç¸é¢é¢ä¿ã¨å æé¢ä¿ã«ãåå注æãã¦ãã ããã
ãµã³ãã«ãµã¤ãºã®è©±ã«é¢ãã¦ã¯åºæ¥ãéã誤ã£ãäºãæ¸ããªãããã«æ³¨æããã¤ããã§ãããç§èªèº«ã®çµ±è¨ãæ¤å®ã«å¯¾ããç解ä¸è¶³ããã説æã表ç¾ã«ééããããããããã¾ããããã®éã¯ææãã¦ããã ããã¨å©ããã¾ãã
åèæ¸ç±
ç§ã仮説æ¤è¨¼ã«åãçµã¿å§ããå½åã確çã¨çµ±è¨ã«é¢ãã¦å ¨ãã®ç¡ç¥ã§ãããæ°å¦ã®ç¥èãä¹ãããåèæ¸ãèªãã§ããªããªãç解ã§ãã¾ããããããªæã«åºä¼ã£ããå ¥éçµ±è¨å¦ãã¨ãä¼ããããã®å¿ççµ±è¨ãã®äºåã¯ãæ··ä¹±ããã¡ãªæãä¸ããä¸å¯§ã«æ¸ããã¦ãããé常ã«å©ããã¾ãããããçµ±è¨ãä¸ããå¦ã¼ãã¨ããæ¹ã«ãããããã¾ããã¾ããä»åã®èª¬æã«ãªããå¹³åå¤ã®æ¤å®ããªã©ãè¡ãéã®ãµã³ãã«ãµã¤ãºã«é¢ãã¦ã¯ãæ°¸ç°éããã®ããµã³ãã«ãµã¤ãºã®æ±ºãæ¹ãã§å¦ã¶äºãã§ãã¾ãã
å ¥é çµ±è¨å¦ âæ¤å®ããå¤å¤é解æã»å®é¨è¨ç»æ³ã¾ã§â - æ å 伸ä¸
https://www.amazon.co.jp/dp/4274068552/
ä¼ããããã®å¿ççµ±è¨: å¹æéã»ä¿¡é ¼åºéã»æ¤å®å - å¤§ä¹ ä¿è¡äº, 岡ç°è¬ä»
https://www.amazon.co.jp/dp/4326250720/
ãµã³ãã«ãµã¤ãºã®æ±ºãæ¹ - æ°¸ç° é
https://www.amazon.co.jp/dp/4254126654/
便å©ãµã¤ã
Optimizely社ããµã³ãã«ãµã¤ãºãè¨ç®ãããã¼ã¸ãç¡åã§æä¾ãã¦ãã¾ã*14ã www.optimizely.com
追è¨
hoxo_mãããæ¬è¨äºã«ä¸è¶³ãã¦ããç¹ãè¨äºã«ãã¦ãã ããã¾ããããããã¨ããããã¾ãã ãææã®ã¨ãããç®æ°ãããã®ã®æ°å¤ã¯çæçã«è¯ããªãå¾åã«ããã¾ããå®åã§ã¯è½ã¡çãã¾ã§ã®ãã¼ã¿ãåãæ¨ã¦ãããã§å¤æãã¦ãã¾ãã
*1:ç´ããããã¦ç³ã訳ããã¾ãããã仮説æ¤è¨¼ã¨ä»®èª¬æ¤å®ã¯å¥ã§ã
*2:å°ãåã§ããå®ã®ã¨ããã¯ç§ãABãã¹ããè¡ãçç±ã¯æ¤å®ã®ããã¨ãããããæ¢åã®æ å ±(ä»åã§è¨ãã°Aã®CVR)ã®å£ç¯å¤åãä»ã®æ½çã®å½±é¿ã観測ãããã¨ã主ç®çã§ããAãä»ã¾ã§10%ã ããã¨ãã£ã¦ãBã®è¨æ¸¬ä¸ã10%ã§ããç¶ããã¨ã¯éãã¾ãããããã¦ã¾ãã解éãå¤æã¯æ¤å®ããããã®èæ¯ã«ããä¿¡é ¼åºéããã¨ã«è¡ã£ã¦ãã¾ããããããªããé©åãªãµã¤ã³ãã«ãµã¤ãºã説æããã«ã¯ä»®èª¬æ¤å®ã®èãã«æ²¿ãæ¹ã容æãªãããä»åã¯ãABãã¹ãã®çµæã仮説æ¤å®ãããæèã§èª¬æãã¾ã
*3:æ£ããã¯CVRã5%ãä½ããå¯è½æ§ãèæ ®ãã両å´æ¤å®ã§ã
*4:å·®ã®å¤§ãããéè¦ãªå ´åã¯ãå·®ã®æç¡ã«å ãã¦ä¿¡é ¼åºéã«ãã解éãå¿ è¦ã¨ãªããå ´åã«ãã£ã¦ã¯ããå¤ãã®ãµã³ãã«ãå¿ è¦ã¨ãã¾ã
*5:ãã ãXæªæºã®å·®ã®æç¡ã«ã¤ãã¦ã¯ä¾ç¶ã¨ãã¦æ å ±ä¸è¶³ã§ã
*6:ä¸è¬çã«ã¯å¹³åå¤ã®å·®ã以ã¦ãå¹æé(Effect size)ãã¨ãã¦èª¬æããã¾ã
*7:ä»åã®èª¬æã§ã¯è§¦ãã¾ããããpå¤ãå©ç¨ãã帰ç¡ä»®èª¬ã®æ£å´ã«ã¤ãã¦ã¯åºæºçã®èª¤ãã«æ³¨æãã¦ãã ãããAlex Reinhartããã®ãThe p value and the base rate fallacy — Statistics Done Wrongã(西åå²æãããæ¥æ¬èªã«ç¿»è¨³ããã¦ãã¾ãã【翻訳】ダメな統計学 (5) p値と基準率の誤り|Colorless Green Ideasã)ãé常ã«ãããããããæ··ä¹±ãããã³ã«èªã¿è¿ããã¦ããã ãã¦ãã¾ã
*8:äºé åå¸ã«ããã確çã®ä¿¡é ¼åºéã¯æ£è¦åå¸ã§è¿ä¼¼ãã¦è¨ç®ãã¦ãã¾ãããã ãããµã³ãã«ãµã¤ãºãå°ãªãå ´ååã³ç¢ºçã極端ã«ä½ã(é«ã)å ´åã®ä¿¡é ¼åºéã«ã¯æ³¨æãå¿ è¦ã§ããæ°ã«ãªãæ¹ã¯Wikipediaã®ã Binomial proportion confidence interval - Wikipediaãããynomuraããã®ã http://ynomura.dip.jp/archives/2010/01/18.htmlãèªãã§ãã ãããå®åã§ã¯Rãªã©ã使ããã¨ããããããã¾ã
*9:αãå°ããã»ã©Î²ã¯å¤§ãããªãã¾ã
*10:1-β(1-0.39)=61%ãæ¤åºå(power)ã¨ããã¾ã
*11:éè§ã«ãªãã¨ãããã¨ã¯ããµã³ãã«ãµã¤ãºã®å¢å ã«ãããå¾ãããçµæãæ¬æ¥ã®å¤ã«è¿ã¥ã確çãå¢ãäºãæå³ãã¦ããããããã大数の法則 - Wikipediaãã¨è¨ãã¾ã
*12:ããªãAã¯2.5%ã§Bã¯5%ãªã®ããAã5%ãããªãã®ããã¨ä¸æè°ã«æãããããã¾ããããããã¯ä¸¡å´æ¤å®ã¨ããèãããã¨ã«Î±ãè¨å®ãã¦ããäºã«é¢ä¿ãã¾ããä»åã®èª¬æã®ç¯å²å¤ã®ãããæ°ã«ãªãæ¹ã¯ã両å´æ¤å®ãããçå´æ¤å®ãã¨ãã£ããã¼ã¯ã¼ãã§èª¿ã¹ã¦ã¿ã¦ãã ãã
*13:äºé åå¸ã®æ£è¦åå¸ã¸ã®è¿ä¼¼ã«ã¯æ¡ä»¶ãããã¾ãã詳ããã¯ã 二項分布 - Wikipediaãã ポアソン分布 - Wikipediaããèªãã§ãã ãã
*14:ä»åã®è¨ç®çµæã¨Optimizely社ã®è¨ç®çµæã¯ç°ãªãã¾ããã©ã®ãããªå¼ãå©ç¨ãã¦ãããã¯åããã¾ããã§ãããããã©ããã«æ¸ãã¦ãããªãã°æãã¦ãã ããã追è¨: æãã¦ããã ãã¾ãã http://pages.optimizely.com/rs/optimizely/images/stats_engine_technical_paper.pdf