A presentation on misleading simplifications and where to find them. We need to name our simplifying assumptions when modeling social scientific data, in order to avoid getting fooled by them. Slide deck available at www.mattiheino.com/naming-assumptions or https://www.slideshare.net/MatinHeino/assumptions-which-must-be-named.
11. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
Raw reality
Simplifying assumptions
Source
12. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
Raw reality
Simplifying assumptions
Source
13. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
Raw reality
Simplifying assumptions
Source
14. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
Raw reality
Simplifying assumptions
Source
15. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
Source
16. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
Source
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
17. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
Source
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
18. www.mattiheino.com
@heinonmatti
Ergodicity
• Time average and ensemble average are equal
• Measuring a single person 1000 times = measuring 1000 people 1 time
• When a process is ergodic, generalising from group-level data to
individuals is ok!
• When non-ergodic, groups and people have different properties
Ecological fallacy
Link for economists
Link for (social) psychologists
19. www.mattiheino.com
@heinonmatti
Empirical study:
• Six samples, compare within- and between individual characteristics
• Positive/negative affect
• Heart rate
• Depressive symptoms
• Means were quite close
• Individuals varied wildly in time Link
28. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
29. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
”What does it mean for my
conclusions, if this assumption
is far from reality?”
30. www.mattiheino.com
@heinonmatti
Perceiving the world
• Draw boundaries between things
• Classify elements into types
• Constructs are different from each other
• Consider structures as unchanging
• All possible events can happen: markov
processes (no memory or hysterisis)
• Only most probable events can happen:
equilibria
• Consider processes as unchanging
• Consider components (people) as
identical and unchanging
Raw reality
Simplifying assumptions
”What does it mean for my
conclusions, if this assumption
is far from reality?”
Link
31. www.mattiheino.com
@heinonmatti
Thanks! Summary:
• Find out (and explicate!) the assumptions you’re making
• Investigate the extent to which they violate reality
• Triangulate with other models and methods when possible
• Always ask: “What does it mean if I’m wrong about this?”
• Next step: Learn about interaction-dominant causality
Slides: mattiheino.com/naming-assumptions