BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//conferences.acspri.org.au//2022//talk//ATW73K
BEGIN:VTIMEZONE
TZID:AEST
BEGIN:STANDARD
DTSTART:20000326T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3;UNTIL=20050326T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20060402T040000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060401T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20070325T040000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=3;UNTIL=20070324T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20080406T040000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000827T030000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=8;UNTIL=20000826T170000Z
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20011028T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20071027T170000Z
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20081005T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=10
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-2022-ATW73K@conferences.acspri.org.au
DTSTART;TZID=AEST:20221123T172000
DTEND;TZID=AEST:20221123T173500
DESCRIPTION:Double Machine Learning (DML) is a causal machine learning meth
 od that promises substantial benefits when estimating average treatment ef
 fects in observational data\, particularly where existing theory is too we
 ak to identify controls or justify a quasi-experimental approach. DML bene
 fits from much of the power and flexibility of predictive models\, while a
 lso giving the unbiased causal estimates of traditional regression approac
 hes. However\, in-practice it often involves relaxing causal identificatio
 n assumptions\, assuming algorithms will correctly identify controls in hi
 gh-dimensional datasets.\n\nExplicitly constraining model fitting with a c
 ausal graph (a diagram laying out which variables cause changes in which) 
 is one solution that has been suggested for better causal identification\,
  but the benefits of this approach are yet to be established and a proper 
 methodology for constructing them has not been laid out (excluding data-dr
 iven causal discovery which has its own serious drawbacks).\n\nThis presen
 tation looks at where DML can be useful in the social sciences and where w
 e might be able to draw in qualitative data to build causal graphs and imp
 rove inference. It covers two studies\, one on returns of education and on
 e on the effect of private schooling on standardised testing performance. 
 In both cases\, causal graphs were constructed by interviewees with varyin
 g levels of background knowledge and models fit under these constraints we
 re compared with unconstrained models. The return on education study used 
 instrumental variables to get yardstick causal estimates for comparison. T
 he private schooling study uses semi-synthetic data to establish ground tr
 uth. While both cases have good existing theory\, relying only on intervie
 ws tests whether it is possible to build pragmatic causal assumptions even
  when theory is poor (as this is often when DML is used).\n\nIn both studi
 es\, constrained and unconstrained DML estimation performed roughly equall
 y well on large samples\, though unconstrained models performed worse on s
 mall subsamples (n = 1000). Importantly\, even a basic level of background
  knowledge outperformed unconstrained DML in these cases. Combining up mul
 tiple graphs into one further reduced bias.\n\nUnconstrained DML seems to 
 be a useful approach where identification is achieved through control-on-o
 bservables and where the sample size is large. However\, a mixed-methods a
 pproach where qualitative data is used to shape causal assumptions may imp
 rove estimation where large samples are not available.\n\nRecording link: 
 https://acspri-org-au.zoom.us/rec/share/fe80_c8_scf5yQJEY_Vv6vJ6R7XFjJUKwZ
 nkpSbPOus_gP1IIKiSQO8FF_LSLDiZ.j0lwiy1bIA1ehAAO?startTime=1669184550000
DTSTAMP:20261007T024911Z
LOCATION:Zoom Breakout Room 2
SUMMARY:Combining qualitative data and causal machine learning for better e
 stimation - Patrick Rehill
URL:https://conferences.acspri.org.au/2022/talk/ATW73K/
END:VEVENT
END:VCALENDAR