What is a Qini Curve 🌱

The Qini Curve is the Cumulative ITE on the treated

  • The area is the area between this curve and the random uplift curve which is just a straight line from (0,0) to the rightmost point of your Qini Curve \(\text { Qini Curve Area }=\sum_{i=0}^{M-1} \frac{1}{2}\left(\frac{n_{t, y=1}\left(\phi_{i+1}\right)-n_{t, y=1}\left(\phi_{i}\right)}{N_{t}}-\frac{n_{c, y=1}\left(\phi_{i+1}\right)-n_{c, y=1}\left(\phi_{i}\right)}{N_{c}}\right) \frac{1}{M}\)

where $\phi_i = i/M$.

\[\text { Qini Curve Area }=\sum_{i=0}^{M-1} \frac{1}{2}\left(\operatorname{Qini} \operatorname{curve}\left(\phi_{i+1}\right)+\operatorname{Qini} \operatorname{curve}\left(\phi_{i}\right)\right)\left(\phi_{i+1}-\phi_{i}\right)\] \[\text { Randomized Qini Area }= \frac{1}{2}(\operatorname{Qini} \operatorname{curve}(\phi_{M}))\] \[Q = \text { Qini Curve Area } - \text { Randomized Qini Area }\] \[f(t)=\left(\frac{Y_{t}^{T}}{N_{t}^{T}}-\frac{Y_{t}^{C}}{N_{t}^{C}}\right)\left(N_{t}^{T}+N_{t}^{C}\right)\] \[g(t)=Y_{t}^{T}-\frac{Y_{t}^{C} N_{t}^{T}}{N_{t}^{C}}\]

for the $t$ datapoints sorted by largest predicted ITE/Uplift, We calculate

  • I think it is just cumulative ITE in order of decreasing predicted ITEs

Notes mentioning this note

There are no notes linking to this note.


Here are all the notes in this garden, along with their links, visualized as a graph.