BEGIN:VCALENDAR
VERSION:2.0
PRODID:talks.ox.ac.uk
BEGIN:VEVENT
SUMMARY:New Results on Minimax Regret Treatment Rules in Finite Samples - 
 Patrik Guggenberger (Penn State University)
DTSTART;VALUE=DATE-TIME:20241115T141500Z
DTEND;VALUE=DATE-TIME:20241115T153000Z
UID:https://talks.ox.ac.uk/talks/id/97538a81-103a-42bd-ade1-0f072923df30/
DESCRIPTION:We study minimax regret treatment rules in finite samples unde
 r matched treatment assignment in a setup where a policymaker\, informed b
 y a sample\, needs to decide between  different treatments for a T≥2. Ra
 ndomized rules are allowed for. We show that the generalization of the min
 imax regret rule derived in Stoye (2009) for the case T = 2 is minimax reg
 ret for general finite T > 2. We also show by example\, that in the case o
 f random assignment the generalization of the minimax rule in Stoye (2009)
  to the case T > 2 is not necessarily minimax regret and derive minimax re
 gret rules for a few small sample cases\, e.g. for N = 2 when T = 3. We al
 so discuss numerical approaches to approximate minimax regret rules for un
 balanced samples. We then study minimax regret treatment rules in finite s
 amples when a specific quantile (rather than expected outcome) is the obje
 ct of interest. We establish that all treatment rules are minimax regret u
 nder ""matched"" and ""random sampling"" schemes while under ""testing an 
 innovation"" no-data rules are shown to be minimax regret.\nSpeakers:\nPat
 rik Guggenberger (Penn State University)
LOCATION:Manor Road Building (Seminar Room C)\, Manor Road OX1 3UQ
TZID:Europe/London
URL:https://talks.ox.ac.uk/talks/id/97538a81-103a-42bd-ade1-0f072923df30/
BEGIN:VALARM
ACTION:display
DESCRIPTION:Talk:New Results on Minimax Regret Treatment Rules in Finite S
 amples - Patrik Guggenberger (Penn State University)
TRIGGER:-PT1H
END:VALARM
END:VEVENT
END:VCALENDAR