Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
統計的因果推論勉強会 第4回
Search
Hikaru Goto
August 27, 2016
Research
0
1.3k
統計的因果推論勉強会 第4回
経営学系統計学エンドユーザーのための統計的因果推論勉強会の第4回目です。
Hikaru Goto
August 27, 2016
Tweet
Share
More Decks by Hikaru Goto
See All by Hikaru Goto
統計的因果推論勉強会第5回
hikaru1122
0
2.3k
R実習 2016年9月25日
hikaru1122
1
2.4k
統計的因果推論勉強会 第3回
hikaru1122
0
1.9k
統計的因果推論勉強会 第2回
hikaru1122
0
1.8k
Other Decks in Research
See All in Research
機械学習と数理最適化の融合-文脈付き確率的最短路を例として-
mickey_kubo
2
510
Alexander Mielke Hellinger--Kantorovich (a.k.a. Wasserstein-Fisher-Rao) Spaces and Gradient Flows
jjzhu
3
200
CASCON 2023 Most Influential Paper Award Talk
tsantalis
0
130
Introduction of NII S. Koyama's Lab (AY2024)
skoyamalab
0
160
Deep State Space Models 101 / Mamba
kurita
9
3.6k
Discovering Universal Geometry in Embeddings with ICA
momoseoyama
1
370
待機電力を削減したネットワーク更新型電子ペーパーサイネージの開発と評価 / IOT64
yumulab
0
110
媒介分析と疫学
kingqwert
0
130
DeepCrysTet: A Deep Learning Approach Using Tetrahedral Mesh for Predicting Properties of Crystalline Materials
tsurubee
0
390
SANER 2019 Most Influential Paper Talk
tsantalis
1
130
Generative Spoken Dialogue Language Modeling [対話論文読み会@電通大]
yuta0306
1
140
Threat Intelligence and Beyond
rishikadesai_7
0
190
Featured
See All Featured
Understanding Cognitive Biases in Performance Measurement
bluesmoon
12
1k
ピンチをチャンスに:未来をつくるプロダクトロードマップ #pmconf2020
aki_iinuma
80
44k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
34
8.9k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
358
22k
Optimizing for Happiness
mojombo
370
69k
10 Git Anti Patterns You Should be Aware of
lemiorhan
649
58k
Web development in the modern age
philhawksworth
203
10k
4 Signs Your Business is Dying
shpigford
176
21k
Raft: Consensus for Rubyists
vanstee
133
6.3k
Easily Structure & Communicate Ideas using Wireframe
afnizarnur
188
16k
Robots, Beer and Maslow
schacon
PRO
155
7.9k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
352
28k
Transcript
ܦӦֶܥ ౷ܭֶΤϯυϢʔβʔͷͨΊͷ ౷ܭతҼՌਪษڧձ ୈ4ճ 20168݄27 @hikaru1122 1
ษڧձͷϞοτʔ • ʮհೖʯ͍ͨ͠ͳΒɺҼՌਪͷํ๏Λʹ͚ͭΑ͏ɻ • ҼՌޮՌΛਪఆ͢Δํ๏ΛֶͿɻ • ֶతͳ͜ͱʹߦ͔ͳ͍ɻ πʔϧͱͯ͠ʹ͚ͭΔɻ 2
ຊͷൣғ • ٶຊɹୈ4ষɹ53ʙ73ท • ຊɹୈ3ষʮIPW ਪఆྔʯ 69ʙ74ท 3
෮श • ࣄ࣮, ճؼੳɼҼՌޮՌɼަབྷɼڞมྔɼڧ͘ແࢹͰ͖Δ ׂΓͯ݅ɼείΞ • ాʢ2014ʣͷୈ16ষʙ17ষ෮शʹͳΔɻ 4
ٶຊɹୈ4ষ • ٶຊͷωλຊͷஶऀʢJudea Pearlʣɻ • 4ষ5ষͷཧతͳ४උͩͱࢥ͏ɻࢲ ͨͪʹେ͖ؔ͘͠ͳ͍ɻ • ͍ͬͯ͏͔ɼαούϦΘ͔Γ·ͤΜɻ 5
Θʔ͍ • ֶతʹ͓ख্͛ɻ • ·ͨڭ͑Λ͍ʹߦ͖·͢ʢʼU༷ʣɻ 6
ٶຊͷ༻ޠͷ֬ೝ • DAG ͰҼՌతҙຯ͚͕ͮՄೳͳͷΛҼՌμΠΞάϥϜͱݺͿ ʢ75ทʣɻ • DAG ͦͷͷ७ਮͳ֬Ϟσϧʢ62ทʣɻ • Ϛϧίϑੑʹۄಥ͖ͷΠϝʔδɻ
લͷঢ়ଶͰ࣍ͷঢ়ଶ͕ܾ·Δʢ;ʔΜʣɻ • ͖݅ಠཱ 7
ͳΜͰ͖݅ಠཱ͕େͳͷʁ • Γ͍ͨͷݪҼม͕݁Ռมʹ༩͑ΔӨڹɻ • ݪҼม݁ՌมʹӨڹΛ͋ͨ͑ΔͷʢަབྷҼࢠʣͷӨڹ ΛऔΓআ͍ͯɼόΠΞεͳ͘ҼՌޮՌΛਪఆ͍ͨ͠ɻ • ަབྷҼࢠͷӨڹΛίϯτϩʔϧͯ͠ɼͦΕͰͳ͓ҼՌޮՌ͕ ೝΊΒΕΔ͔Ͳ͏͔ΛΓ͍ͨɻ 8
ͦ͜ͰࠓίϨɻ • Pearl, Glymour and Jewell(2016) • ҼՌϞσϧσʔλ͕ੜ͞ΕΔϝΧ χζϜͰ͋Δɻ 9
ҼՌμΠΞάϥϜͷجຊ3ύλʔϯ • ࿈ʢchainʣɼذʢforkʣɼ߹ྲྀʢcolliderʣ • ͲΕ͕ͲΕʹӨڹΛ༩͍͑ͯΔ͔ΠϝʔδͰ͖ΕOKͰɻ • ࣍ճʢٶຊ ୈ5ষʣͷཧղʹͱͯେࣄ 10
࿈ʢchainʣ • Xֶ͕ߍͷࢿۚɼY͕ςετͷɼZ͕߹֨ • YΛҰఆͳʹ੍ݶͰ͖ΔͳΒɼXͱZͲΜͳΛͱͬͯ OKʢ͖݅ಠཱʣ 11
ذʢforkʣ • X͕ؾԹɼY͕ΞΠεΫϦʔϜച্ɼZ͕൜ࡑ • YͱZ͚ͩݟͨΒɼ૬͕ؔ͋ΔͧɻY͔ΒZͷҼՌޮՌ͋Δͷ͔ ͳʁ 12
ذʢforkʣ • YͱZʹٖ૬ؔͷڪΕ͕͋Δɻ • XΛ੍ݶͯ͠ɼYͱZͷ૬ؔΛݟΕΑ͍ɻ • ͏͜ͷखͷେৎͰ͢Ͷʂ 13
߹ྲྀʢcolliderʣ • Pearl͞ΜΒʹΑΔͱͱͬͯେࣄͳܗΒ͍͠ʢextremely important to the study of causalityʣ 14
߹ྲྀʢcolliderʣ • ͢Ͱʹࢲͨͪ͜ͷة͏͞Λ͍ͬͯ·͢ʢલճͬͨʣɻ • ZΛҰఆʹͨ͠ΒɼXͱYʹ૬͕ؔੜͯ͡͠·͏ɻ 15
ࠓͷٶຊ·ͱΊ • ҼՌͷߏΛਤͰॻ͘ͱɼ͝རӹ͋Δ ͔Αɻ • ෮शͱͯ͠ɼؠσʔλαΠΤϯεୈ3 רͷ28ʙ38ทʢྛɾࠇ 2016ʣ͕ײ ಈతʹΘ͔Γ͍͢ɻ •
͞Βʹ39ʙ46ทΛಡΜͰ͓͘ͱɼ࣍ճ ͷେࣄͳͱ͜Ζ͕ཧղͰ͖ΔʢͨͿ Μʣɻ 16
ຊɹୈ3ষʮIPWਪఆྔʯ • ΑΓΑ͘ҼՌޮՌͷਪఆΛܭࢉͰ͖Δํ๏ɻ • ࠓ࣍ͷެ͚ࣜͩͰOKɻ ɹ,ɹ • 2ͭΛҾ͖ࢉ͢ΕΑ͍ɻ 17
Ͳ͜Λܭࢉ͍ͯ͠Δͷ͔ʁ • ATEɿʢᶃʴᶄʣͷฏۉ ʔʢᶅʴᶆʣͷฏۉ • ATTͷ߹ผͷࣜʹͳΔɻ • ͜ΕҎ্ઌʢDRਪఆྔʣʹਐΈ·ͤΜɻ 18
IPWਪఆྔɹܭࢉͷ࣮ࡍ • ʮҼՌޮՌͷਪఆʂRͰ࣮ફ - είΞɼϚονϯάɼIPW ਪఆྔ -ʯ • ʮؠDS3αϙʔτϖʔδʯ •
ʮؠσʔλαΠΤϯεvol.3ͷσʔλͰ༡΅͏ʯ • 3ͭΊ͔ͳΓRʹ׳Εͯͳ͍ͱ͍͠ɻ • 1ͭΊͱ2ͭΊ͕ཧղͰ͖ΔΑ͏ʹͳΖ͏ɻ 19
IPWਪఆྔΛ༻͍ͨจ • ຊޠͰগͳ͍ɻӳޠͰະௐࠪɻ • ࠓ࣍ͷ2ຊɻ • ಛఆอ݈ࢦಋͷ༧հೖࢪࡦͷޮՌʹؔ͢ΔݚڀʢੴΒ 2013ʣ • ࣾձతݽཱͱϥΠϑΠϕϯτͷؔ࿈ʢࡾ୩
2015ʣ 20
ੴΒʢ2013ʣ SAS༻ • ʮϝλϘରࡦʢ݈ͱอ݈ࢦಋʣͬͯΈͨʯ • ʮड͚ͨਓͱड͚ͳ͔ͬͨਓͷҧ͍Λݟ͍ͨʯ • ʮ͍Ζ͍ΖվળͰ͖ͨΑʂʯ • ड͚ͨਓ924ਓɼड͚ͳ͔ͬͨਓ3128ਓ
• ϚονϯάΛͨ͠Βଟ͘ͷσʔλΛࣺͯΔ͜ͱʹͳΔͷͰIPWਪ ఆྔΛ༻͍ͨͷ͔ͳʁ 21
ࡾ୩ʢ2015ʣ ༻ιϑτෆ໌ • ʮࣾձతݽཱʹͲͷΑ͏ͳϥΠϑΠϕϯτ͕ؔ࿈͔ͨ͠ΛΓ ͍ͨʯ • ʮWebௐࠪैདྷͷํ๏ΑΓόΠΞε͕͋Δ͔Βௐ͍ͨ͠ʯ • ʮͦ͜Ͱٯ֬ॏΈ͚๏Λ͓͏ʂʯ •
ͻΐͬͱͯ͠ɼຊୈ6ষͷ༰Ͱʂʁ 22
JGSS 23
JGSS 24
࣍ճ༧ࠂ • ٶຊ ୈ5ষɹόοΫυΞج४ • ຊ ୈ4ষɹ4.1ʙ4.3 ͱ 4.7 •
ؠDS vol.3 ͷ28ʙ48ทɼ62ʙ90ท͕ಡΈ͍͢Ͱ͢ɻ • ςΫχΧϧͳʹ໎͍ࠐΉҰาखલ͔ʂʁ • ܦӦֶͷҰྲྀδϟʔφϧSMJʹܝࡌ͞Εͨ౷ܭతҼՌਪͷߟ ΛಡΜͰɼཱͪҐஔΛ֬ೝɻ 25
ࢀߟจݙ • Pearl, J., Glymour, M. and Jewell, N. P.
(2016). Causal Inference in Statistics: A Primer. John Wiley & Sons. • ੴળथ, ࠓҪതٱ, தඌ༟೭, ᜊ౻૱, ా٢࣏. (2013). ಛఆอ݈ࢦಋͷ༧հೖࢪࡦͷޮՌʹؔ͢Δݚڀ: େنσʔλϕʔεΛ༻ͨ͠είΞʹΑΔҼՌੳ. ް ੜͷࢦඪ, 60(5), 1-6. 26
ࢀߟจݙ • খౡོɾࢁຊক࢙(2013). ExcelͰֶͿڞࢄߏੳͱ άϥϑΟΧϧϞσϦϯά. ΦʔϜࣾ. • ྛַɾࠇֶ(2016). ૬ؔͱҼՌͱؙͱҹͷͳ͠ɼؠ σʔλαΠΤϯεɼvol.3ɼ28ʙ48.
27
ࢀߟจݙ • ྛޫ. (2012). JGSS ౷ܭੳηϛφʔ 2011-είΞɾ ΣΠςΟϯά๏Λ༻͍ΔҼՌੳ. ຊ൛૯߹తࣾձௐࠪڞ ಉݚڀڌݚڀจू,
(12), 107ʙ127. • ਸ(2010). ௐࠪ؍σʔλͷ౷ܭՊֶɹҼՌਪɾબ όΠΞεɾσʔλ༥߹. ؠॻళ. 28
ࢀߟจݙ • ਸ(2016). ౷ܭతҼՌޮՌͷجૅɼؠσʔλαΠΤϯ εɼvol.3ɼ62ʙ90. • ࡾ୩ΔΑ. (2015). ࣾձతݽཱͱϥΠϑΠϕϯτͷؔ࿈: είΞ๏ʹΑΔ
Web ௐࠪσʔλੳ͔Β. ཾ୩େֶࣾձ ֶ෦لཁ= Bulletin of the Faculty of Sociology, Ryukoku University, (47), 58-69. 29
ࢀߟจݙ • ٶխາ(2004). ౷ܭతҼՌਪʔճؼੳͷ৽͍͠Έ ʔ. ேॻళ. 30