Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up for free
G-methods for time-varying treatments (Causal inference: What if, Chapter 21-1)
Shuntaro Sato
November 25, 2020
Science
0
1.2k
G-methods for time-varying treatments (Causal inference: What if, Chapter 21-1)
Keywords: 因果推論, Time-varying, G-formula, IP weighting, Doubly robust estimation
Shuntaro Sato
November 25, 2020
Tweet
Share
More Decks by Shuntaro Sato
See All by Shuntaro Sato
「回帰分析から分かること」と「変数選択」
shuntaros
4
5.8k
対照群がない研究デザインで効果を推定する(時系列分断デザイン・自己対照研究デザイン)
shuntaros
2
3k
自己対照デザイン:ケースクロスオーバーデザイン・ケースタイムコントロールデザイン
shuntaros
1
810
何が知りたいのか?〜どのぐらい?に答える〜(医学統計学・疫学セミナー)
shuntaros
0
1.1k
何が知りたいのか?〜因果推論:効果の定義、 Estimandの前振り〜(医学統計学・疫学セミナー)
shuntaros
1
1.1k
何が知りたいのか?〜Estimand、傾向スコアマッチングは何を見てる?〜(医学統計学・疫学セミナー)
shuntaros
1
1.2k
Run in periodと決定木と生存時間解析(MERC勉強会)
shuntaros
0
150
Effect modification(Causal inference: What if, Chapter 4)
shuntaros
0
2.3k
IP Weighting and Marginal Structural Models(Causal inference: What if, Chapter 12)
shuntaros
0
1.9k
Other Decks in Science
See All in Science
深層学習による自然言語処理 輪読会#1 資料
tok41
0
270
2020年京都大学_統計力学
kamakiri1225
0
280
2020年京都大学_量子力学
kamakiri1225
0
370
Теория байесовских сетей - осень 2021 - 2 лекция
dscs
0
110
不審なURLの見つけ方
secchick
1
180
[勉強会資料メモ] Double/Debiased ML
masa_asa
0
110
セミパラメトリックアプローチによる因果探索
sshimizu2006
0
110
The likelihood function and its discontents
dwhgg
0
860
B.LEAGUE におけるバスケットボールのリアルタイム勝利確率モデルの構築 / Realtime win probability model for B.LEAGUE
konakalab
0
120
Astronomical software
dfm
1
520
2020年京都大学_電磁気学
kamakiri1225
0
530
Accumulated Local Effects(ALE)で機械学習モデルを解釈する / TokyoR95
dropout009
2
2.7k
Featured
See All Featured
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
498
130k
Fireside Chat
paigeccino
11
1.2k
Designing with Data
zakiwarfel
91
3.9k
Rebuilding a faster, lazier Slack
samanthasiow
62
7.2k
The Brand Is Dead. Long Live the Brand.
mthomps
45
2.7k
Build your cross-platform service in a week with App Engine
jlugia
219
17k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
237
19k
Automating Front-end Workflow
addyosmani
1351
200k
Principles of Awesome APIs and How to Build Them.
keavy
113
15k
JazzCon 2018 Closing Keynote - Leadership for the Reluctant Leader
reverentgeek
172
8.3k
Building an army of robots
kneath
299
40k
Why Our Code Smells
bkeepers
PRO
324
54k
Transcript
None
・G-methods for time-fixed treatments 本日の内容 ・The g-formula for time-varying treatments
・IP weighting for time-varying treatments ・A doubly robust estimator for time-varying treatments
・G-methods for time-fixed treatments 本日の内容 ・The g-formula for time-varying treatments
・IP weighting for time-varying treatments ・A doubly robust estimator for time-varying treatments
Stratification effect measure modification (-) effect measure modification (+) Mantel-Haenszel
method 別々にオッズ比を報告
Why model? effect measure modification (-) effect measure modification (+)
別々にオッズ比を報告(1つの効果を報告できない) g-methods
g-formula A=1を代入 A=0を代入
IP weighting marginal structural model
Conditional or Marginal? outcome regression saturated parametric stratification g-formula IP
weighting g-estimation or algebraically equivalent
Time-varying treatment g-methods
・G-methods for time-fixed treatments 本日の内容 ・The g-formula for time-varying treatments
・IP weighting for time-varying treatments ・A doubly robust estimator for time-varying treatments
前提 ・本章ではidentifiability conditions(sequential exchangeability, positivity, and consistency)のviolationが ないものとする。 ・static treatment
strategies (always treat vs. never treat) の効果を推定する。
g-formula (weighted average) ・time-fixed treatment (A1 の反実アウトカム) ・time-varying treatment
g-formula (weighted average)
g-formula (weighted average)
g-formula (simulation) のシミュレーション と
g-formulaの注意点 ・DAGに基づいたcovariates L1 をモデルに含める ・static sequential exchangeabilityが成立すればstatic treatment strategyの効果はidentify可能
g-formulaの一般化 ・static treatment strategy ・dynamic treatment strategy linear regression logistic
regression
・G-methods for time-fixed treatments 本日の内容 ・The g-formula for time-varying treatments
・IP weighting for time-varying treatments ・A doubly robust estimator for time-varying treatments
IP weighting (weights) ・nonstabilized IP weights ・ stabilized IP weights
IP weighting (non-stabilized)
Stabilized weights non-stabilized weights: stabilized weights: Lと独立であればよい Lと独立であればよい
IP weighting (stabilized)
IP weightingの一般化 ・nonstabilized IP weights ・ stabilized IP weights logistic
regression logistic regression (misspecifiedでも可)
Marginal Structural Model ・2K > Nのときは推定できない ・marginal structural mean model
stabilized IP weightsを使って推定 misspecified??
Effect Measure Modification ・baseline variable VによるEMMがある場合、marginal structural modelは以下の通り(parametric) stabilized IP
weightsを使って推定 Vに入れて良いのはbaseline variableだけ
・G-methods for time-fixed treatments 本日の内容 ・The g-formula for time-varying treatments
・IP weighting for time-varying treatments ・A doubly robust estimator for time-varying treatments
Doubly Robust Estimator ・g-formula ・ IP weighting
1. Doubly Robust (time-fixed) 2. 3. A=1とA=0でそれぞれ を推定 を推定 ,
をLについて標準化
1. Doubly Robust (time-varying) 2. 3. を推定 からパラメータ を求める。 を求めておく
を推定し、Aの値に応じた を求める。 これを繰り返して を求める。 always treat
Discussion