diff --git a/paper/sl.tex b/paper/sl.tex index 00014b4583f6910e21c294b00274f906b3332a19..42be2104fc66a5c372e502c91a081c1260951859 100755 --- a/paper/sl.tex +++ b/paper/sl.tex @@ -33,7 +33,7 @@ \usepackage{footnote} % show footnotes in tables \makesavenoteenv{table} - +\newcommand{\antti}[1]{{{\color{orange} [AH: #1]}}} \newcommand{\ourtitle}{A Causal Approach for Selective Labels} @@ -168,6 +168,8 @@ We wish to calculate the probability of undesired outcome (\outcome = 0) at a fi & = \sum_\featuresValue \prob{\outcome = 0 | \decision = 1, \features = \featuresValue} \prob{\decision = 1 | \leniency = \leniencyValue, \features = \featuresValue} \prob{\features = \featuresValue} \end{align*} +\antti{Here one can drop do even at the first line according to do-calculus rule 2, i.e. $P(Y=0|do(R=r))=P(Y=0|R=r)$. However, do-calculus formulas should be computed by first learning a graphical model and then computing the marginals using the graphical model. This gives more accurate result. Michael's complicated formula essentially does this, including forcing $P(Y=0|T=0,X)=0$ (the model supports context-specific independence $Y \perp X |�T=0$.)} + Expanding the above derivation for model \score{\featuresValue} learned from the data \[ \score{\featuresValue} = \prob{\outcome = 0 | \features = \featuresValue, \decision = 1}, @@ -221,7 +223,7 @@ random variables so that \prob{\outcome = 0| \features = \featuresValue} = \dfrac{1}{1+\exp\{-\featuresValue\}}. \] -The decision variable $\decision$ was set to 0 if the probability $\prob{\outcome = 0| \features = \featuresValue}$ resided in the top $(1-\leniencyValue)\cdot 100 \%$ of the subjects appointed for that judge. +The decision variable $\decision$ was set to 0 if the probability $\prob{\outcome = 0| \features = \featuresValue}$ resided in the top $(1-\leniencyValue)\cdot 100 \%$ of the subjects appointed for that judge. \antti{How was the final Y determined? I assume $Y=1$ if $T=0$, if $T=1$ $Y$ was randomly sampled from $\prob{\outcome| \features = \featuresValue}$ above? Delete this comment when handled.} Results for estimating the causal quantity $\prob{\outcome = 0 | \doop{\leniency = \leniencyValue}}$ with various levels of leniency $\leniencyValue$ under this model are presented in Figure \ref{fig:without_unobservables}. \begin{figure}