SlideShare a Scribd company logo
1 of 17
Download to read offline
Can a variable be a cause if it
doesn’t correspond with an
intervention?
Can we disagree about philosophy but agree about epidemiology?
1
Session motivated by the “No Manipulation-
No Causation” debate
• “Effects” are the difference in potential outcomes under two alternative
exposures or intervention strategies
• Therefore, impossible to define an effect of something that cannot be intervened
upon
• Easy question: where will I wake up if I take the blue pill instead of the red pill?
• Hard question: would I go fishing more often if I were a man?
Session motivated by the “No Manipulation-
No Causation” debate
• “Effects” are the difference in potential outcomes under two alternative
exposures or intervention strategies
• Therefore, impossible to define an effect of something that cannot be intervened
upon
• Easy question: where will I wake up if I take the blue pill instead of the red pill?
• Hard question: would I go fishing more often if I were a man?
3
Session motivated by the “No Manipulation-
No Causation” debate
• “Effects” are the difference in potential outcomes under two alternative
exposures or intervention strategies
• Therefore, impossible to define an effect of something that cannot be intervened
upon
• Easy question: where will I wake up if I take the blue pill instead of the red pill?
• Hard question: would I go fishing more often if I were a man? A black man?
4
NO Manipulation-NO Causation Perspectives
•Strict NOMNOC (Holland- the father of NOMNOC):
• “Causes are experiences that units undergo and not attributes
that they possess”
• Ruled out age, sex, or race as causes
•Flexible NOMNOC: okay to think of
implausible/impossible interventions
• As long as the intervention does not “fundamentally” change the person.
• “actions” are causes, “states” are not causes: exercise regimen is a
coherent cause, BMI is not (Taubman and Hernan)
• NOT NOMNOC: Manipulation is not necessary, or if you
permit implausible interventions, not much is ruled out
• Sex is randomized at conception. It’s easy to assess the
effects.
• Ignore NOMNOC criterion: Who cares? let’s just talk
about the “mediators”.
5
Social epidemiologists love
to argue about NOMNOC
•Race doesn’t affect your health- racism affects health!
•Race doesn’t affect your health- changing your race
would make you a whole different person.
6
Social epidemiologists love
to argue about NOMNOC
•Race doesn’t affect your health- racism affects health!
•Treating race as a cause does not reify race, or deny the
obvious facts that “race” is fluidly defined within social
contexts and social context shapes the effects of race on
health.
•Treating race as a cause does not privilege genetic or
social mechanisms for racial inequalities. It merely opens
these up for rigorous evaluation.
•Race doesn’t affect your health- changing your race
would make you a whole different person.
•What defines me a person? My eye color? My hair color?
My parents’ SES? This is an impossible rabbit hole.
7
Why the NOMNOC perspective is (often) useful
• Specifying an intervention that corresponds with the exposure
variable clarifies what you mean and helps guide public health
actions
• Income = Wages? Lottery winnings? Welfare?
• Physical activity = Marathons? Boxing? Wheelchair stretches?
• This is a huge advantage, and it has more practical, day to day implications for
how we do research than the disadvantages
• Although in theory either mistake -- to fail to articulate the corresponding
intervention OR to ignore important non-manipulable risk factors – is
relevant…
• ….in practice, failing to articulate the corresponding intervention is more
common and more pernicious.
• MG’s perspective: People who deny race/sex are causes usually analyze data
as if they believed race/sex were causes
• (VanderWeele and Robinson 2014 provided a coherent way for them to talk about their regressions)8
Why the NOMNOC perspective is damaging
• Using causal language allows access to an array of useful
machinery for causal inference for many important questions:
• DAGs, d-separation rules, recognition of collider bias
• How do you select control variables? How do you interpret adjusted regression
coefficients?
• Do you control for income when estimating the effect of sex on depression?
• NOMNOC handicaps researchers interested in racial and gender
disparities, because it is unclear how to adopt and use causal
inference framework.
• If we can’t talk about race and sex as causes, discussing
inequalities and the mechanisms by which race and sex
influence health is a tongue twister and fails to communicate
• The effect of race on CVD is mediated by…
Becomes
• The non-causal correlation between race and CVD is attenuated by
adjusting for… 9
Why the NOMNOC perspective is damaging
• If you cannot conceptualize “states” as causes, you cannot
generalize from one intervention to variations on that
intervention that would have the same effects
• Reducing blood pressure with drug A reduces MI risk. Maybe other drugs that
reduce blood pressure would also reduce MI risk.
• It is knowledge of the mediating state between an intervention and an outcome
that guides development of new interventions
10
Mandatory
schooling
Lower college
tuition
Universal
pre-K
Dementia Years of
schooling
Mandatory
schooling
Lower college
tuition
Universal
pre-K
Dementia
Miguel perspective [slides deleted]
11
What we disagree on
• MG: Strict NOMNOC is a handicap. Flexible NOMNOC is largely harmless.
• MG: You can apply d-separation, causal inference machinery and thinking while
treating race, sex, and states as causes.
• MG: refusing to define sex or race as a cause makes it difficult to take advantage
of the tools for rigorously evaluating the origins of inequalities
• MH: Ok, let’s say race is a cause. How does that help us interpret the number
we obtain from data when comparing health outcomes in, say, blacks and
whites? How does it help us decide what to do next? Whatever you answered to
the last question, wouldn’t you do the same next thing after observing the
white/black differences even if you didn’t say “race is a cause”?
• Is it useful to conceptualize “states” rather than “actions” as causal?
• MG we learn generalizable information by defining states as causes and investigating
• MH: Wait, wouldn’t we equally investigate after finding differences between different
states even if we don’t say “this state is a cause”?
• MG: Sure, you don’t have to say it, but wouldn’t it be easier to use words to describe
this idea?
12
What we agree on
• It is usually helpful to articulate an intervention that would correspond
with the exposure of interest
• Exceptions: descriptions of inequality when the “cause” is not of interest as a
target of intervention (e.g., race or sex) but rather the mechanisms
• Interventions do not need to be humanly feasible
• Interventions on certain variables are ambiguous and usually entail
more description to be clear: if we intervene on “race” do we also
intervene on location of birth, to match the geographic distribution of
births of the other race?
• Many “state” variables correspond with many different possible
interventions on distinct underlying variables, which may all have
different consequences when wiggled.
• Racial and gender inequalities in health are deplorable, modifiable, and
we should seek interventions to eliminate such inequalities.
13
Maria’s response
14
The argument against states as causes:
Distinguishing “fat hand” problems from
consistency problems
• Fat hand interventions are interventions that
influence more than just the specific exposure
of interest
• The intervening hand was too fat to just grab
the precise exposure of interest
• As in pickup sticks, the fat hand touches lots of
sticks and has many extraneous consequences
• For some exposures, all known interventions
may be “fat hand”
• BMI: diet, exercise, surgery…
15
Distinguishing “fat hand” problems from
consistency problems
• Fat hand interventions influence more than
just the specific exposure of interest
• Consistency violations reflect a measurement
that is too crude: exposures with different
outcomes are called the same thing, ie the
exposure is heterogeneous
• BMI: subcutaneous adipose tissue, visceral
adipose tissue
Consistency violations make it difficult to predict the consequences of future
interventions, because you do not know which flavor of the exposure has
which effect. 16
Neither fat hands nor inconsistency preclude
causation
• Adiposity and many other physiologic states have both fat hand problems and
consistency problems
• This doesn’t mean they are not “causes”
• Consistency violations occur because we do not fully understand or have not
measured the relevant aspect of the exposure. Better research may allow us
clarify the specific exposure of interest.
• Fat hand interventions may not be a problem, depending what else is being
triggered by the intervention.
• Fat hand problems and consistency violations make it difficult to draw causal
inferences… but they don’t imply the exposure (or some flavor of the exposure)
is not causal
17

More Related Content

Similar to Miguel maria causation_mg to share2

Chapter 1What do the authors mean by sexual intelligence” Be.docx
Chapter 1What do the authors mean by sexual intelligence” Be.docxChapter 1What do the authors mean by sexual intelligence” Be.docx
Chapter 1What do the authors mean by sexual intelligence” Be.docx
zebadiahsummers
 
Gender workshop for physicians.pptx
Gender workshop for physicians.pptxGender workshop for physicians.pptx
Gender workshop for physicians.pptx
chcjayanagara
 
Question #1 Please choose ONE of the following to make an origi.docx
Question #1 Please choose ONE of the following to make an origi.docxQuestion #1 Please choose ONE of the following to make an origi.docx
Question #1 Please choose ONE of the following to make an origi.docx
simonlbentley59018
 
Hope and Action: Patient Interviewing Tips
Hope and Action: Patient Interviewing TipsHope and Action: Patient Interviewing Tips
Hope and Action: Patient Interviewing Tips
taralv
 

Similar to Miguel maria causation_mg to share2 (20)

HLEG thematic workshop on "Multidimensional Subjective Well-being", Angus Deaton
HLEG thematic workshop on "Multidimensional Subjective Well-being", Angus DeatonHLEG thematic workshop on "Multidimensional Subjective Well-being", Angus Deaton
HLEG thematic workshop on "Multidimensional Subjective Well-being", Angus Deaton
 
Reflecting on mental health consumer-survivor-expatient movement
Reflecting on mental health consumer-survivor-expatient movementReflecting on mental health consumer-survivor-expatient movement
Reflecting on mental health consumer-survivor-expatient movement
 
Session 8 a iariw discussion wolfson paper -ruggles
Session 8 a iariw discussion wolfson paper -rugglesSession 8 a iariw discussion wolfson paper -ruggles
Session 8 a iariw discussion wolfson paper -ruggles
 
NEW UROP LAKE S
NEW UROP LAKE SNEW UROP LAKE S
NEW UROP LAKE S
 
8th grade get real
8th grade get real8th grade get real
8th grade get real
 
History taking in Psychosexual Medicine
History taking in Psychosexual MedicineHistory taking in Psychosexual Medicine
History taking in Psychosexual Medicine
 
How to Have Difficult Conversations: Notes Nov 2015
How to Have Difficult Conversations: Notes Nov 2015How to Have Difficult Conversations: Notes Nov 2015
How to Have Difficult Conversations: Notes Nov 2015
 
Chapter 1What do the authors mean by sexual intelligence” Be.docx
Chapter 1What do the authors mean by sexual intelligence” Be.docxChapter 1What do the authors mean by sexual intelligence” Be.docx
Chapter 1What do the authors mean by sexual intelligence” Be.docx
 
Gender workshop for physicians.pptx
Gender workshop for physicians.pptxGender workshop for physicians.pptx
Gender workshop for physicians.pptx
 
Rhodes College student lecture 31.05.16
Rhodes College student lecture 31.05.16Rhodes College student lecture 31.05.16
Rhodes College student lecture 31.05.16
 
11a bolte san antoniofinal2010bolte
11a bolte san antoniofinal2010bolte11a bolte san antoniofinal2010bolte
11a bolte san antoniofinal2010bolte
 
Webinar: What Did I Miss? The Hidden Costs of Depriortizing Diversity in User...
Webinar: What Did I Miss? The Hidden Costs of Depriortizing Diversity in User...Webinar: What Did I Miss? The Hidden Costs of Depriortizing Diversity in User...
Webinar: What Did I Miss? The Hidden Costs of Depriortizing Diversity in User...
 
O Behave! Issue 18
O Behave! Issue 18O Behave! Issue 18
O Behave! Issue 18
 
Treatment Concepts and Techniques in Sexual Therapy
Treatment Concepts and Techniques in Sexual TherapyTreatment Concepts and Techniques in Sexual Therapy
Treatment Concepts and Techniques in Sexual Therapy
 
Sexual Health & Colorectal Cancer - May 2018 Webinar
Sexual Health & Colorectal Cancer - May 2018 WebinarSexual Health & Colorectal Cancer - May 2018 Webinar
Sexual Health & Colorectal Cancer - May 2018 Webinar
 
L Catterall & D Middleton - Sexual dysfunction for women
L Catterall & D Middleton - Sexual dysfunction for womenL Catterall & D Middleton - Sexual dysfunction for women
L Catterall & D Middleton - Sexual dysfunction for women
 
Question #1 Please choose ONE of the following to make an origi.docx
Question #1 Please choose ONE of the following to make an origi.docxQuestion #1 Please choose ONE of the following to make an origi.docx
Question #1 Please choose ONE of the following to make an origi.docx
 
Hope and Action: Patient Interviewing Tips
Hope and Action: Patient Interviewing TipsHope and Action: Patient Interviewing Tips
Hope and Action: Patient Interviewing Tips
 
Addressing Multiculturalism in Health Care Presentation
Addressing Multiculturalism in Health Care PresentationAddressing Multiculturalism in Health Care Presentation
Addressing Multiculturalism in Health Care Presentation
 
Bias in Healthcare: An Evidence-Based Overview
Bias in Healthcare: An Evidence-Based OverviewBias in Healthcare: An Evidence-Based Overview
Bias in Healthcare: An Evidence-Based Overview
 

Recently uploaded

會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
中 央社
 
Spellings Wk 4 and Wk 5 for Grade 4 at CAPS
Spellings Wk 4 and Wk 5 for Grade 4 at CAPSSpellings Wk 4 and Wk 5 for Grade 4 at CAPS
Spellings Wk 4 and Wk 5 for Grade 4 at CAPS
AnaAcapella
 
Personalisation of Education by AI and Big Data - Lourdes Guàrdia
Personalisation of Education by AI and Big Data - Lourdes GuàrdiaPersonalisation of Education by AI and Big Data - Lourdes Guàrdia
Personalisation of Education by AI and Big Data - Lourdes Guàrdia
EADTU
 
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
EADTU
 

Recently uploaded (20)

Andreas Schleicher presents at the launch of What does child empowerment mean...
Andreas Schleicher presents at the launch of What does child empowerment mean...Andreas Schleicher presents at the launch of What does child empowerment mean...
Andreas Schleicher presents at the launch of What does child empowerment mean...
 
Improved Approval Flow in Odoo 17 Studio App
Improved Approval Flow in Odoo 17 Studio AppImproved Approval Flow in Odoo 17 Studio App
Improved Approval Flow in Odoo 17 Studio App
 
Trauma-Informed Leadership - Five Practical Principles
Trauma-Informed Leadership - Five Practical PrinciplesTrauma-Informed Leadership - Five Practical Principles
Trauma-Informed Leadership - Five Practical Principles
 
UChicago CMSC 23320 - The Best Commit Messages of 2024
UChicago CMSC 23320 - The Best Commit Messages of 2024UChicago CMSC 23320 - The Best Commit Messages of 2024
UChicago CMSC 23320 - The Best Commit Messages of 2024
 
Spring gala 2024 photo slideshow - Celebrating School-Community Partnerships
Spring gala 2024 photo slideshow - Celebrating School-Community PartnershipsSpring gala 2024 photo slideshow - Celebrating School-Community Partnerships
Spring gala 2024 photo slideshow - Celebrating School-Community Partnerships
 
ANTI PARKISON DRUGS.pptx
ANTI         PARKISON          DRUGS.pptxANTI         PARKISON          DRUGS.pptx
ANTI PARKISON DRUGS.pptx
 
OSCM Unit 2_Operations Processes & Systems
OSCM Unit 2_Operations Processes & SystemsOSCM Unit 2_Operations Processes & Systems
OSCM Unit 2_Operations Processes & Systems
 
Mattingly "AI and Prompt Design: LLMs with NER"
Mattingly "AI and Prompt Design: LLMs with NER"Mattingly "AI and Prompt Design: LLMs with NER"
Mattingly "AI and Prompt Design: LLMs with NER"
 
Book Review of Run For Your Life Powerpoint
Book Review of Run For Your Life PowerpointBook Review of Run For Your Life Powerpoint
Book Review of Run For Your Life Powerpoint
 
AIM of Education-Teachers Training-2024.ppt
AIM of Education-Teachers Training-2024.pptAIM of Education-Teachers Training-2024.ppt
AIM of Education-Teachers Training-2024.ppt
 
8 Tips for Effective Working Capital Management
8 Tips for Effective Working Capital Management8 Tips for Effective Working Capital Management
8 Tips for Effective Working Capital Management
 
Observing-Correct-Grammar-in-Making-Definitions.pptx
Observing-Correct-Grammar-in-Making-Definitions.pptxObserving-Correct-Grammar-in-Making-Definitions.pptx
Observing-Correct-Grammar-in-Making-Definitions.pptx
 
會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文會考英文
 
Mattingly "AI & Prompt Design: Named Entity Recognition"
Mattingly "AI & Prompt Design: Named Entity Recognition"Mattingly "AI & Prompt Design: Named Entity Recognition"
Mattingly "AI & Prompt Design: Named Entity Recognition"
 
24 ĐỀ THAM KHẢO KÌ THI TUYỂN SINH VÀO LỚP 10 MÔN TIẾNG ANH SỞ GIÁO DỤC HẢI DƯ...
24 ĐỀ THAM KHẢO KÌ THI TUYỂN SINH VÀO LỚP 10 MÔN TIẾNG ANH SỞ GIÁO DỤC HẢI DƯ...24 ĐỀ THAM KHẢO KÌ THI TUYỂN SINH VÀO LỚP 10 MÔN TIẾNG ANH SỞ GIÁO DỤC HẢI DƯ...
24 ĐỀ THAM KHẢO KÌ THI TUYỂN SINH VÀO LỚP 10 MÔN TIẾNG ANH SỞ GIÁO DỤC HẢI DƯ...
 
An Overview of the Odoo 17 Knowledge App
An Overview of the Odoo 17 Knowledge AppAn Overview of the Odoo 17 Knowledge App
An Overview of the Odoo 17 Knowledge App
 
Spellings Wk 4 and Wk 5 for Grade 4 at CAPS
Spellings Wk 4 and Wk 5 for Grade 4 at CAPSSpellings Wk 4 and Wk 5 for Grade 4 at CAPS
Spellings Wk 4 and Wk 5 for Grade 4 at CAPS
 
Personalisation of Education by AI and Big Data - Lourdes Guàrdia
Personalisation of Education by AI and Big Data - Lourdes GuàrdiaPersonalisation of Education by AI and Big Data - Lourdes Guàrdia
Personalisation of Education by AI and Big Data - Lourdes Guàrdia
 
VAMOS CUIDAR DO NOSSO PLANETA! .
VAMOS CUIDAR DO NOSSO PLANETA!                    .VAMOS CUIDAR DO NOSSO PLANETA!                    .
VAMOS CUIDAR DO NOSSO PLANETA! .
 
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
Transparency, Recognition and the role of eSealing - Ildiko Mazar and Koen No...
 

Miguel maria causation_mg to share2

  • 1. Can a variable be a cause if it doesn’t correspond with an intervention? Can we disagree about philosophy but agree about epidemiology? 1
  • 2. Session motivated by the “No Manipulation- No Causation” debate • “Effects” are the difference in potential outcomes under two alternative exposures or intervention strategies • Therefore, impossible to define an effect of something that cannot be intervened upon • Easy question: where will I wake up if I take the blue pill instead of the red pill? • Hard question: would I go fishing more often if I were a man?
  • 3. Session motivated by the “No Manipulation- No Causation” debate • “Effects” are the difference in potential outcomes under two alternative exposures or intervention strategies • Therefore, impossible to define an effect of something that cannot be intervened upon • Easy question: where will I wake up if I take the blue pill instead of the red pill? • Hard question: would I go fishing more often if I were a man? 3
  • 4. Session motivated by the “No Manipulation- No Causation” debate • “Effects” are the difference in potential outcomes under two alternative exposures or intervention strategies • Therefore, impossible to define an effect of something that cannot be intervened upon • Easy question: where will I wake up if I take the blue pill instead of the red pill? • Hard question: would I go fishing more often if I were a man? A black man? 4
  • 5. NO Manipulation-NO Causation Perspectives •Strict NOMNOC (Holland- the father of NOMNOC): • “Causes are experiences that units undergo and not attributes that they possess” • Ruled out age, sex, or race as causes •Flexible NOMNOC: okay to think of implausible/impossible interventions • As long as the intervention does not “fundamentally” change the person. • “actions” are causes, “states” are not causes: exercise regimen is a coherent cause, BMI is not (Taubman and Hernan) • NOT NOMNOC: Manipulation is not necessary, or if you permit implausible interventions, not much is ruled out • Sex is randomized at conception. It’s easy to assess the effects. • Ignore NOMNOC criterion: Who cares? let’s just talk about the “mediators”. 5
  • 6. Social epidemiologists love to argue about NOMNOC •Race doesn’t affect your health- racism affects health! •Race doesn’t affect your health- changing your race would make you a whole different person. 6
  • 7. Social epidemiologists love to argue about NOMNOC •Race doesn’t affect your health- racism affects health! •Treating race as a cause does not reify race, or deny the obvious facts that “race” is fluidly defined within social contexts and social context shapes the effects of race on health. •Treating race as a cause does not privilege genetic or social mechanisms for racial inequalities. It merely opens these up for rigorous evaluation. •Race doesn’t affect your health- changing your race would make you a whole different person. •What defines me a person? My eye color? My hair color? My parents’ SES? This is an impossible rabbit hole. 7
  • 8. Why the NOMNOC perspective is (often) useful • Specifying an intervention that corresponds with the exposure variable clarifies what you mean and helps guide public health actions • Income = Wages? Lottery winnings? Welfare? • Physical activity = Marathons? Boxing? Wheelchair stretches? • This is a huge advantage, and it has more practical, day to day implications for how we do research than the disadvantages • Although in theory either mistake -- to fail to articulate the corresponding intervention OR to ignore important non-manipulable risk factors – is relevant… • ….in practice, failing to articulate the corresponding intervention is more common and more pernicious. • MG’s perspective: People who deny race/sex are causes usually analyze data as if they believed race/sex were causes • (VanderWeele and Robinson 2014 provided a coherent way for them to talk about their regressions)8
  • 9. Why the NOMNOC perspective is damaging • Using causal language allows access to an array of useful machinery for causal inference for many important questions: • DAGs, d-separation rules, recognition of collider bias • How do you select control variables? How do you interpret adjusted regression coefficients? • Do you control for income when estimating the effect of sex on depression? • NOMNOC handicaps researchers interested in racial and gender disparities, because it is unclear how to adopt and use causal inference framework. • If we can’t talk about race and sex as causes, discussing inequalities and the mechanisms by which race and sex influence health is a tongue twister and fails to communicate • The effect of race on CVD is mediated by… Becomes • The non-causal correlation between race and CVD is attenuated by adjusting for… 9
  • 10. Why the NOMNOC perspective is damaging • If you cannot conceptualize “states” as causes, you cannot generalize from one intervention to variations on that intervention that would have the same effects • Reducing blood pressure with drug A reduces MI risk. Maybe other drugs that reduce blood pressure would also reduce MI risk. • It is knowledge of the mediating state between an intervention and an outcome that guides development of new interventions 10 Mandatory schooling Lower college tuition Universal pre-K Dementia Years of schooling Mandatory schooling Lower college tuition Universal pre-K Dementia
  • 12. What we disagree on • MG: Strict NOMNOC is a handicap. Flexible NOMNOC is largely harmless. • MG: You can apply d-separation, causal inference machinery and thinking while treating race, sex, and states as causes. • MG: refusing to define sex or race as a cause makes it difficult to take advantage of the tools for rigorously evaluating the origins of inequalities • MH: Ok, let’s say race is a cause. How does that help us interpret the number we obtain from data when comparing health outcomes in, say, blacks and whites? How does it help us decide what to do next? Whatever you answered to the last question, wouldn’t you do the same next thing after observing the white/black differences even if you didn’t say “race is a cause”? • Is it useful to conceptualize “states” rather than “actions” as causal? • MG we learn generalizable information by defining states as causes and investigating • MH: Wait, wouldn’t we equally investigate after finding differences between different states even if we don’t say “this state is a cause”? • MG: Sure, you don’t have to say it, but wouldn’t it be easier to use words to describe this idea? 12
  • 13. What we agree on • It is usually helpful to articulate an intervention that would correspond with the exposure of interest • Exceptions: descriptions of inequality when the “cause” is not of interest as a target of intervention (e.g., race or sex) but rather the mechanisms • Interventions do not need to be humanly feasible • Interventions on certain variables are ambiguous and usually entail more description to be clear: if we intervene on “race” do we also intervene on location of birth, to match the geographic distribution of births of the other race? • Many “state” variables correspond with many different possible interventions on distinct underlying variables, which may all have different consequences when wiggled. • Racial and gender inequalities in health are deplorable, modifiable, and we should seek interventions to eliminate such inequalities. 13
  • 15. The argument against states as causes: Distinguishing “fat hand” problems from consistency problems • Fat hand interventions are interventions that influence more than just the specific exposure of interest • The intervening hand was too fat to just grab the precise exposure of interest • As in pickup sticks, the fat hand touches lots of sticks and has many extraneous consequences • For some exposures, all known interventions may be “fat hand” • BMI: diet, exercise, surgery… 15
  • 16. Distinguishing “fat hand” problems from consistency problems • Fat hand interventions influence more than just the specific exposure of interest • Consistency violations reflect a measurement that is too crude: exposures with different outcomes are called the same thing, ie the exposure is heterogeneous • BMI: subcutaneous adipose tissue, visceral adipose tissue Consistency violations make it difficult to predict the consequences of future interventions, because you do not know which flavor of the exposure has which effect. 16
  • 17. Neither fat hands nor inconsistency preclude causation • Adiposity and many other physiologic states have both fat hand problems and consistency problems • This doesn’t mean they are not “causes” • Consistency violations occur because we do not fully understand or have not measured the relevant aspect of the exposure. Better research may allow us clarify the specific exposure of interest. • Fat hand interventions may not be a problem, depending what else is being triggered by the intervention. • Fat hand problems and consistency violations make it difficult to draw causal inferences… but they don’t imply the exposure (or some flavor of the exposure) is not causal 17