Presentation is loading. Please wait.

Presentation is loading. Please wait.

Safer science: making psychological science more Transparent & replicable Simine Vazire UC Davis.

Similar presentations


Presentation on theme: "Safer science: making psychological science more Transparent & replicable Simine Vazire UC Davis."— Presentation transcript:

1 Safer science: making psychological science more Transparent & replicable
Simine Vazire UC Davis

2 Make predictions and test them. Show your work.
What is science? Make predictions and test them. Show your work.

3 overview PART I: Values and norms in science PART II: Stats review PART III: Crisis PART IV: Solutions

4 Part i: values and norms
PART I: Values and norms in science PART II: Stats review PART III: Crisis PART IV: Solutions

5 Values and norms in science
What makes science different from other ways of knowing? What makes science different from pseudoscience? What are the core values of science?

6 Part i: values and norms
What makes science different from other ways of knowing?

7 Ways to Acquire Knowledge
Intuition: I feel that it is so Authority: A trusted source says that it is so Empiricism: I’ll believe it when I see it Rationalism: Logically, it follows that… { Scientific Knowledge Authority – you get a parking ticket. Someone in authority is telling you a truth.

8 Part i: values and norms
What makes science different from other ways of knowing? Focus on empiricism and rationality What makes science different from pseudoscience?

9 What makes science different from pseudoscience?
What is “the scientific method”? Positivism: Base conclusions on objective facts Problem: No such thing (always requires some interpretation) Falsificationism: Make predictions that can be wrong Problem: All scientific theories are wrong Successive approximation: Make predictions that are closer and closer to reality Problem: How do we know if we’re getting closer? Maybe there is no such thing as THE scientific method?

10 Some “rules” for science
Make predictions and test them. Show your work.

11 Part i: values and norms
Instead of THE scientific method, there are norms and values that are followed in science. This helps distinguish science from pseudoscience Robert Merton, a sociologist, studied the “ethos” of science Came up with four “norms” that help define science Can also help distinguish science from pseudoscience

12 Merton’s norms Universalism Communality Disinterestedness
The validity of a scientific claim does not depend on who is making it. No hierarchy. Status should not matter. Communality The findings of science belong to everyone, they are not private property. No secrecy. Open communication is key. Disinterestedness Scientists should be focused on finding the truth, not on their own success. No self-interest. Report whatever you find, even if it makes you look bad. Organized skepticism Do not take things at face value. Verify others’ claims. Nothing is sacred.

13 “What we call scientific knowledge today is a body of statements of varying degrees of certainty. Some of them are most unsure; some of them are nearly sure; but none is absolutely certain. Scientists are used to this. We know that it is consistent to be able to live and not know. Some people say, ‘How can you live without knowing?’ I do not know what they mean. I always live without knowing. That is easy. How you get to know is what I want to know.” Organized skepticism

14 Transparency as a core value
“The credibility of scientific claims comes, in part, from the fact that their meaning is, at a minimum, available for other scholars to rigorously evaluate […] Such open access to the origins of others’ claims is the hallmark of scientific ways of knowing. Accordingly, when social scientists fail to document their assumptions, decisions, and actions […] it limits others’ ability to understand the meaning of scientists’ claims.” -Lupia & Elman, 2014, p. 20

15 Transparency leads to reproducible science
“Nullius in Verba” Take no one’s word Science means never having to say “Trust me” Others should be able to reproduce my findings

16 The ethos of science The overarching goal of science is to get closer and closer to the truth We do this by finding out what is reproducible, and throwing out what is not Self-correction is an active process, and requires (Merton’s norms): No hierarchy – anyone can challenge anyone else’s finding Transparency – everything is out in the open No conflicts of interest – truth is more important than personal success Skepticism – tearing things down is as valuable as finding new things For Merton, these were moral values, not just habits Scientists do not always (ever?) completely live up to these values

17 Self-correction in science

18 Some “RULES” for science
Make predictions and test them. Show your work.

19 Part i: values and norms
What makes science different from other ways of knowing? Focus on empiricism and rationality What makes science different from pseudoscience? Testable predictions, falsifiability, verifiability What are the core values of science? Transparency, reproducibility, self-correction

20 So I have just one wish for you—the good luck to be somewhere where you are free to maintain the kind of integrity I have described, and where you do not feel forced by a need to maintain your position in the organization, or financial support, or so on, to lose your integrity.  May you have that freedom. -Richard Feynman

21 Part Ii: stats review PART I: Values and norms in science PART II: Stats review PART III: Crisis PART IV: Solutions

22 Errors in Science

23 not significant p < .05 Is there an effect? True state of the world
No effect Effect No effect not significant Our Conclusion p < .05 Effect

24 not significant p < .05 Is there an effect? True state of the world
No effect Effect Correct Rejection No effect not significant Our Conclusion Hit p < .05 Effect

25 not significant p < .05 Two kinds of errors True state of the world
No effect Effect Correct Rejection False Negative No effect not significant Our Conclusion Hit p < .05 Effect

26 not significant p < .05 Two kinds of errors True state of the world
No effect Effect Correct Rejection False Negative No effect not significant Our Conclusion False Positive Hit p < .05 Effect

27 False positive rate True state of the world Correct Rejection False
No effect Effect Correct Rejection False Negative No effect 95% Our Conclusion False Positive 5% ??% Hit ??% Effect

28 So what is the % of false positives?

29 Part IiI: CRisis PART I: Values and norms in science PART II: Stats review PART III: Crisis PART IV: Solutions

30 97% 37% xx Open Science Collaboration, 2015, Science

31

32

33 How to turn anything into a significant result
Did you get the effect you predicted? No Yes Did you get ANY effect? HARK! Yes Publish

34 HARKing

35

36 Make predictions and test them. Show your work.
What is science? Make predictions and test them. Show your work. [The same dataset cannot be used to generate the hypothesis and to test it]

37 How to turn anything into a significant result
Did you get the effect you predicted? No Yes Did you get ANY effect? HARK! Yes No Publish p-hack! Can you dig around and find one? Yes

38 Source: Leif Nelson’s slides

39 H1: The U.S. economy is affected by whether Republicans or Democrats are in office

40 H1: The U.S. economy is affected by whether Republicans or Democrats are in office

41

42 How to turn anything into a significant result
Did you get the effect you predicted? No Yes Did you get ANY effect? HARK! Yes No Publish p-hack! Can you dig around and find one? Yes No File drawer

43 % significant (Psychology)
Fanelli (2012)

44 False positive rate True state of the world Correct Rejection False
No effect Effect Correct Rejection False Negative No effect 95% Our Conclusion False Positive 5% Hit Effect

45

46 Make predictions and test them. Show your work.
Where we went wrong Make predictions and test them. Show your work.

47 Why not just do science right?
We didn’t know better. It’s hard.

48

49 What is wrong with people?
“The first principle is that you must not fool yourself – and you are the easiest person to fool.” -Richard Feynman

50 What is wrong with people?
Students start out with the right ideas about science That you should make predictions ahead of time That you should try to prove yourself wrong That you should report everything you find Incentive structure encourages cutting corners Overconfidence leads us to believe incredible results Scientific community rewards scientists who can produce amazing results

51 Part IV: solutions PART I: Values and norms in science PART II: Stats review PART III: Crisis PART IV: Solutions

52 How can we do better?

53 When is a finding likely to be true?
Large sample

54 Why are large samples better?

55 When is a finding likely to be true?
Large sample Small p-value (well below p < .05)

56 Why p = .04 should be rare

57 Why p = .04 should be rare

58 When is a finding likely to be true?
Large sample Small p-value (well below p < .05) Pre-registered

59 What is pre-registration?
Source: Leif Nelson’s slides

60 When is a finding likely to be true?
Large sample Small p-value (well below p < .05) Pre-registered Full disclosure (transparency)

61 What is transparency? Making materials publicly available Sharing data
Disclosing flexibility in data analysis If you have a prediction, pre-register If you don’t, don’t HARK

62 Why transparency? “The credibility of scientific claims comes, in part, from the fact that their meaning is, at a minimum, available for other scholars to rigorously evaluate […] Such open access to the origins of others’ claims is the hallmark of scientific ways of knowing. Accordingly, when social scientists fail to document their assumptions, decisions, and actions […] it limits others’ ability to understand the meaning of scientists’ claims.” -Lupia & Elman, 2014, p. 20

63 Why transparency?

64 When is a finding likely to be true?
Large sample Small p-value (well below p < .05) Pre-registered Full disclosure (transparency) Has been replicated

65

66 When is a finding likely to be true?
Large sample Small p-value (well below p < .05) Pre-registered Full disclosure (transparency) Has been replicated* Believable

67 Dreber et al., 2015, PNAS

68 Make predictions and test them. Show your work.
How can we do better? Make predictions and test them. Show your work.

69

70 Science is hard www.improvingpsych.org
To make science more scientific, we need : More transparency More accountability Lower expectations

71 The end


Download ppt "Safer science: making psychological science more Transparent & replicable Simine Vazire UC Davis."

Similar presentations


Ads by Google