SlideShare uma empresa Scribd logo
1 de 19
Reconsidering Baron and Kenny:
Myths and Truths
about Mediation Analysis
Xinshu Zhao, John G. Lynch Jr., and Qimei Chen (JCR, 2010)
Minhwan Lee
2013 (Best Article for 2010), JCR Best Article Award Winners
Introduction
 Misapplication of the Baron and Kenny
 Data do not comport with one or more of the Baron and Kenny criteria
 1) cause authors to drop projects that may be promising
 2) cause journals to reject paper that may deserve publication
 Misunderstanding of mediation: Ignore important hints for theory building
 Baron and Kenny (1986): cited by 44,155, June 2014, Google scholar
 While the popularity of the Baron-Kenny procedure continues to grow, a small
technical literature has grown alongside showing flaws in Baron and Kenny’s logic
 Fitzsimons (2008) and Irwin and McClelland (2001) translated an existing
technical literature on moderated regression for practicing consumer researchers
 Similarly, we aim to explain to users of Baron and Kenny’s tests how new
developments should change how they test for mediation
-2-
Baron and Kenny’s Tests
 1) Baron and Kenny claim that mediation is strongest when there is an indirect
effect but no direct effect in equation 3
 2) there need not be a significant “effect to be mediated” in equation 2.
 3) the Sobel test is low in power compared to a bootstrap test popularized by
Preacher and Hayes (2004), in some cases markedly so.
-3-
Mediators Hidden in “Direct” Effects: Boon to Theory Building
 Baron and Kenny (1986)
 “full mediation”: indirect effect but no direct effect
 “partial mediation”: both indirect and direct effects
 Iacobucci (2008, 12): “when all tests are properly conducted and reported, the
majority of articles conclude with ‘partial mediation.”  mediation is usually
accompanied by a direct effect
 The concept of a “direct” effect is clear statistically, but it is often unclear
theoretically
 Bolton, Cohen, and Bloom’s (2006), Mitra and Lynch (1995)
 “direct” paths often result from omission of one or more mediators from the
model (Shrout and Bolger 2002):
 there is a silver lining in “partial mediation.”: The sign of the mysterious “direct”
effect has heuristic value for theory building.
-4-
No Need for an “Effect to be mediated”
 Competitive and complementary mediations are equally likely and of equal
theoretical interest a priori
 c and a×b are of the same sign: complementary mediation
 c and a×b are of the opposite sign: competitive mediation
 only complementary mediations should be judged to be publishable
 Hypothetical modification of Mitra and Lynch’s (1995)
 the unexplained negative direct path to deter publishing findings of a positive
indirect path
 (a) advertising increases consideration set size, which in turn increases price
sensitivity; (b) despite the progress in the paper, there is room for future work
accounting for the direct effect; and (c) in searching for added mediators, future
authors should focus first on those that would produce a negative indirect path
 Focused discussion would be considerably more helpful than most “Future
Research” sections now published
-5-
A Typology of Mediations and Nonmediations
 three patterns consistent with mediation and two with nonmediation
 complementary mediation: “consistent” model or “positive confounding”
 competitive mediation: “inconsistent” model or “negative confounding”
 Proper interpretation of one’s data requires two dimensions for the indirect
path and the direct path
-6-
Five Typology (Suggested) Baron and Kenny
1
Complementary mediation: Mediated effect (a×b) and direct effect (c)
both exist and point at the same direction
overlaps with
partial mediation
2
Competitive mediation: Mediated effect (a×b) and direct effect (c)
both exist and point in opposite directions
no mediation
3 Indirect-only mediation: Mediated effect (a×b) exists, but no direct effect
overlaps with
full mediation
4 Direct-only nonmediation: Direct effect (c) exists, but no indirect effect no mediation
5 No-effect nonmediation: Neither direct effect nor indirect effect exists no mediation
A Typology of Mediations and Nonmediations
 In our approach to mediation analysis, c’ now represents only the total effect—
not the “effect to be mediated.”
 The X-Y test is never relevant to establishing mediation
 be possible to establish an indirect effect despite no total effect
 Figure 2*: Decision Tree for Establishing and Understanding Types of
Mediation and Nonmediation
 decision tree to conceptualize these five types of mediation and nonmediation to
convey to readers what really matters in a mediation analysis
 The top of the figure (2a) shows the statistical path to establishing mediation and
classifying its type
 The bottom of the figure (2b) shows the interpretation of the data pattern for
conclusions about theory
-7-
A Typology of Mediations and Nonmediations
-8-
Baron and Kenny’s three equations are useful;
Sobel test, by a superior bootstrap test
(Preacher and Hayes 2004, 2008)
The main role for
Baron and Kenny’s
three equations is
in deciding the type of
mediation
A Typology of Mediations and Nonmediations
-9-
all three cases on the left, the data support
the hypothesized mediation story X→M→Y
For both complementary and competitive mediations, the significant
direct effect c points to the possible existence of some omitted second
mediator that can be pursued in future research
the sign of the direct effect can point
to as yet undiscovered mediators
Sobel’s Not Noble
 Preacher and Hayes (2004) made a strong case that it is insufficient to show
that the effect of X on Y is reduced in size when M is added to the model
 the main contribution of Preacher and Hayes (2004) was to present SAS and
SPSS syntax for an alternative “bootstrap” test of the indirect effect that is
almost always more powerful than Sobel’s test
 95% confidence interval will often improperly include zero, compared to the 95%
confidence interval we would create if we could observe the sampling distribution
of a×b
 Zhao (1997): collected during Super Bowl 1994 to show how to perform
Preacher-Hayes bootstrap tests in either SAS or SPSS
 http://www.comm.hkbu.edu.hk/zhao/shared
-10-
Sobel’s Not Noble
 To perform bootstrapping in SPSS, open your data set and then follow the steps
below:
 i. Open the Preacher-Hayes script in SPSS for Windows (.sbs file).
 ii. Run the Preacher-Hayes script from the scripting window to activate the dialog box.
 iii. Identify your dependent (i.e., liking), independent (frequency), and one or more
mediating variables (here only clutter). You may also identify covariate variables if
appropriate.
 iv. Set bootstrap samples to, say, 5,000.
 v. Set confidence level at 95% or 99%.
 vi. Click OK to execute the script.
 vii. Find “Bootstrap Results for Indirect Effects” from the output window and the 95%
confidence interval (-.0930 to -.0268 in our case). If the confidence interval does not
include 0, the indirect effect a×b is significant and mediation is established, which takes
you to the left of figure 2. If the confidence interval includes 0, a×b is not significant and
mediation hypothesis is rejected, which takes you to the right of figure 2.
-11-
Sobel’s Not Noble
-12-
 Open data: 1994 Super Bowl Data SPSS  Analyze- Regression - Preacher and Hayes
(2008) Multiple Mediation (INDIRECT)
Sobel’s Not Noble
-13-
 Results (Consistent with Figure 3)
Sobel’s Not Noble
-14-
 Results (Consistent with Figure 3)
Recommended Steps for Mediation Analysis
 the indirect effect is significant
 Simply run the Preacher Hayes script and generate “Bootstrap Results for Indirect
Effects,” as we showed in figure 3, to determine whether the indirect effect a×b is
significant and thus whether to take the left (mediation) or right (nonmediation)
branch of figure 2
 Use SEM to estimate all parameters simultaneously, or you may run the two
regression equations 1 and 3; Preacher-Hayes script output (fig. 3) provides
these estimates automatically under “Direct and Total Effects.”
 i. If a×b is significant but c is not, you have indirect only mediation.
 ii. If a×b is not significant but c is, you have direct only nonmediation.
 iii. If neither a×b nor c is significant, you have no effect nonmediation.
 iv. If both a×b and c are significant, determine the sign of a×b×c by multiplying the
three coefficients, or by multiplying c by the mean value of a×b from the bootstrap
output. If a×b×c is positive, it is complementary mediation; if a×b×c is negative, it
is competitive mediation.
-15-
Recommended Steps for Mediation Analysis
 Report also the unstandardized regression coefficients a, b, and c to allow
substantive interpretation of the results
-16-
• a×b=-.0527
• 95% confidence interval
excluding zero (-.0930 to -.0268)
• a=2.53
• b=-.0208
• c=.057
• a×b×c=-.0030;
negative/competitive mediation
Unexpectedly Flipped Signs of Indirect Paths
 Cohen and Cohen (1975) and Friedman and Wall (2005) note that an indirect
effect opposite in sign to the zero-order correlations arises in cases of “net
suppression”
 All cases of net suppression in mediation analysis can be seen as special cases of
competitive mediation in which the direct path dominates the indirect
 To avoid publishing a conclusion that is wrong, authors should report (and be
asked by journals to report) the type of mediation and actual a, b, and c
coefficients, not just the significance test of the indirect effect
-17-
Structural Equation Models and Baron and Kenny’s Regression Approach
 Iacobucci (2008) argues that structural equation (SEM) approaches dominate
the “causal steps” approach of Baron and Kenny (1986)
 SEM approach is superior to Baron and Kenny’s because it estimates everything
simultaneously instead of assuming that equations 1–3 are independent
 1) even if there is no “true” direct effect or omitted second mediator, one can
observe a significant direct effect due to measurement error in the indicator of
M(Birnbaum and Mellers 1979)
 2) sometimes a researcher’s data may seem to conform to Baron and Kenny’s
conditions for mediation, but the “mediator” is not conceptually different from
the independent variable: it is effectively a manipulation check
-18-
Conclusion
 We have simplified these criticisms and added our own, providing an
overarching framework that considers two dimensions—the indirect effect and
the direct effect—rather than the one dimensional “full,” “partial,” and “none”
classification employed by Baron and Kenny.
 1) although Baron and Kenny and current practice hold up “full mediation” as
the gold standard, most studies report “partial mediation” with a significant
direct path c
 2) there need not be a significant rXY in a proper mediation analysis; The only
requirement for mediation is that the indirect effect a×b be significant
 3) in testing the significance of an indirect effect a×b, use the more rigorous
and powerful bootstrap test, not Sobel
-19-

Mais conteúdo relacionado

Mais procurados

Test-development-final-REPPORT.pptx
Test-development-final-REPPORT.pptxTest-development-final-REPPORT.pptx
Test-development-final-REPPORT.pptx
EzriCoda1
 
Single subjects-research
Single subjects-researchSingle subjects-research
Single subjects-research
Noor Hasmida
 
Prosocial Behavior
Prosocial BehaviorProsocial Behavior
Prosocial Behavior
NikhilRai62
 

Mais procurados (20)

Attribution ppt
Attribution pptAttribution ppt
Attribution ppt
 
Test-development-final-REPPORT.pptx
Test-development-final-REPPORT.pptxTest-development-final-REPPORT.pptx
Test-development-final-REPPORT.pptx
 
Neuropsychological Assessment
Neuropsychological AssessmentNeuropsychological Assessment
Neuropsychological Assessment
 
Social cognition
Social cognitionSocial cognition
Social cognition
 
Attraction - Social Psychology
Attraction - Social PsychologyAttraction - Social Psychology
Attraction - Social Psychology
 
Social perception
Social perceptionSocial perception
Social perception
 
Social cognition
Social cognitionSocial cognition
Social cognition
 
Marital Discord and Divorce By Dr. Ashok Balsekar
Marital Discord and Divorce By Dr. Ashok BalsekarMarital Discord and Divorce By Dr. Ashok Balsekar
Marital Discord and Divorce By Dr. Ashok Balsekar
 
Positive Psychology
Positive PsychologyPositive Psychology
Positive Psychology
 
Moderation and Mediation | Dissertation Webinar
Moderation and Mediation | Dissertation WebinarModeration and Mediation | Dissertation Webinar
Moderation and Mediation | Dissertation Webinar
 
Single subjects-research
Single subjects-researchSingle subjects-research
Single subjects-research
 
Male and female brain what constitutes the gender
Male and female brain  what constitutes the genderMale and female brain  what constitutes the gender
Male and female brain what constitutes the gender
 
Prosocial Behaviour
Prosocial BehaviourProsocial Behaviour
Prosocial Behaviour
 
Null hypothesis for a Factorial ANOVA
Null hypothesis for a Factorial ANOVANull hypothesis for a Factorial ANOVA
Null hypothesis for a Factorial ANOVA
 
SDCT
SDCTSDCT
SDCT
 
Role of Family Pathology....Bhupendra singh
Role of Family Pathology....Bhupendra singhRole of Family Pathology....Bhupendra singh
Role of Family Pathology....Bhupendra singh
 
types of attachment styles
types of attachment stylestypes of attachment styles
types of attachment styles
 
WESCHLERS ADULT INTELLIGENCE SCALE IV
WESCHLERS ADULT INTELLIGENCE SCALE IVWESCHLERS ADULT INTELLIGENCE SCALE IV
WESCHLERS ADULT INTELLIGENCE SCALE IV
 
Prosocial Behavior
Prosocial BehaviorProsocial Behavior
Prosocial Behavior
 
Couple Therapy
Couple Therapy Couple Therapy
Couple Therapy
 

Semelhante a Reconsidering baron and kenny

EDCO 745Regression Data Output And Writeup Grading RubricCrit
EDCO 745Regression Data Output And Writeup Grading RubricCritEDCO 745Regression Data Output And Writeup Grading RubricCrit
EDCO 745Regression Data Output And Writeup Grading RubricCrit
EvonCanales257
 
You clearly understand the concepts of this assignment. You’ve don.docx
You clearly understand the concepts of this assignment. You’ve don.docxYou clearly understand the concepts of this assignment. You’ve don.docx
You clearly understand the concepts of this assignment. You’ve don.docx
jeffevans62972
 
Read the article Competition or Complement Six Sigma and TOC that.docx
Read the article Competition or Complement Six Sigma and TOC that.docxRead the article Competition or Complement Six Sigma and TOC that.docx
Read the article Competition or Complement Six Sigma and TOC that.docx
makdul
 
Statistics 106Homework 6Due Feb. 17, 2016, In Class.docx
Statistics 106Homework 6Due  Feb. 17, 2016, In Class.docxStatistics 106Homework 6Due  Feb. 17, 2016, In Class.docx
Statistics 106Homework 6Due Feb. 17, 2016, In Class.docx
dessiechisomjj4
 
LEARNING OUTCOMESKnow what descriptive statistics are an.docx
LEARNING OUTCOMESKnow what descriptive statistics are an.docxLEARNING OUTCOMESKnow what descriptive statistics are an.docx
LEARNING OUTCOMESKnow what descriptive statistics are an.docx
smile790243
 
MLR Project (Onion)
MLR Project (Onion)MLR Project (Onion)
MLR Project (Onion)
Chawal Ukesh
 
Error Control and Severity
Error Control and SeverityError Control and Severity
Error Control and Severity
jemille6
 
QNT 561 FINAL EXAMS
QNT 561 FINAL EXAMSQNT 561 FINAL EXAMS
QNT 561 FINAL EXAMS
gradings
 
Res 342 final exam
Res 342 final examRes 342 final exam
Res 342 final exam
mn8676766
 
Res 342 final exam
Res 342 final examRes 342 final exam
Res 342 final exam
nbvyut9878
 
[GeertvanKollenburg]-firstyearpaper
[GeertvanKollenburg]-firstyearpaper[GeertvanKollenburg]-firstyearpaper
[GeertvanKollenburg]-firstyearpaper
Geert van Kollenburg
 

Semelhante a Reconsidering baron and kenny (20)

Testing Mediation and regression analysis
Testing Mediation and regression analysisTesting Mediation and regression analysis
Testing Mediation and regression analysis
 
Mediation Analysis in research
Mediation Analysis in researchMediation Analysis in research
Mediation Analysis in research
 
EDCO 745Regression Data Output And Writeup Grading RubricCrit
EDCO 745Regression Data Output And Writeup Grading RubricCritEDCO 745Regression Data Output And Writeup Grading RubricCrit
EDCO 745Regression Data Output And Writeup Grading RubricCrit
 
You clearly understand the concepts of this assignment. You’ve don.docx
You clearly understand the concepts of this assignment. You’ve don.docxYou clearly understand the concepts of this assignment. You’ve don.docx
You clearly understand the concepts of this assignment. You’ve don.docx
 
Representing and generating uncertainty effectively presentatıon
Representing and generating uncertainty effectively presentatıonRepresenting and generating uncertainty effectively presentatıon
Representing and generating uncertainty effectively presentatıon
 
Read the article Competition or Complement Six Sigma and TOC that.docx
Read the article Competition or Complement Six Sigma and TOC that.docxRead the article Competition or Complement Six Sigma and TOC that.docx
Read the article Competition or Complement Six Sigma and TOC that.docx
 
Statistics 106Homework 6Due Feb. 17, 2016, In Class.docx
Statistics 106Homework 6Due  Feb. 17, 2016, In Class.docxStatistics 106Homework 6Due  Feb. 17, 2016, In Class.docx
Statistics 106Homework 6Due Feb. 17, 2016, In Class.docx
 
Hypothesis Testing: Central Tendency – Non-Normal (Compare 1:Standard)
Hypothesis Testing: Central Tendency – Non-Normal (Compare 1:Standard)Hypothesis Testing: Central Tendency – Non-Normal (Compare 1:Standard)
Hypothesis Testing: Central Tendency – Non-Normal (Compare 1:Standard)
 
LEARNING OUTCOMESKnow what descriptive statistics are an.docx
LEARNING OUTCOMESKnow what descriptive statistics are an.docxLEARNING OUTCOMESKnow what descriptive statistics are an.docx
LEARNING OUTCOMESKnow what descriptive statistics are an.docx
 
Simple-Mediation-Analysis.pptx
Simple-Mediation-Analysis.pptxSimple-Mediation-Analysis.pptx
Simple-Mediation-Analysis.pptx
 
MLR Project (Onion)
MLR Project (Onion)MLR Project (Onion)
MLR Project (Onion)
 
Error Control and Severity
Error Control and SeverityError Control and Severity
Error Control and Severity
 
QNT 561 FINAL EXAMS
QNT 561 FINAL EXAMSQNT 561 FINAL EXAMS
QNT 561 FINAL EXAMS
 
Qnt 561 final exams
Qnt 561 final examsQnt 561 final exams
Qnt 561 final exams
 
Hypothesis Testing: Central Tendency – Non-Normal (Compare 2+ Factors)
Hypothesis Testing: Central Tendency – Non-Normal (Compare 2+ Factors)Hypothesis Testing: Central Tendency – Non-Normal (Compare 2+ Factors)
Hypothesis Testing: Central Tendency – Non-Normal (Compare 2+ Factors)
 
Res 342 final exam
Res 342 final examRes 342 final exam
Res 342 final exam
 
Res 342 final exam
Res 342 final examRes 342 final exam
Res 342 final exam
 
Qnt 561
Qnt 561Qnt 561
Qnt 561
 
Econometrics
EconometricsEconometrics
Econometrics
 
[GeertvanKollenburg]-firstyearpaper
[GeertvanKollenburg]-firstyearpaper[GeertvanKollenburg]-firstyearpaper
[GeertvanKollenburg]-firstyearpaper
 

Último

An Overview of Mutual Funds Bcom Project.pdf
An Overview of Mutual Funds Bcom Project.pdfAn Overview of Mutual Funds Bcom Project.pdf
An Overview of Mutual Funds Bcom Project.pdf
SanaAli374401
 
1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdf
QucHHunhnh
 
Seal of Good Local Governance (SGLG) 2024Final.pptx
Seal of Good Local Governance (SGLG) 2024Final.pptxSeal of Good Local Governance (SGLG) 2024Final.pptx
Seal of Good Local Governance (SGLG) 2024Final.pptx
negromaestrong
 
Gardella_PRCampaignConclusion Pitch Letter
Gardella_PRCampaignConclusion Pitch LetterGardella_PRCampaignConclusion Pitch Letter
Gardella_PRCampaignConclusion Pitch Letter
MateoGardella
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
ciinovamais
 
Gardella_Mateo_IntellectualProperty.pdf.
Gardella_Mateo_IntellectualProperty.pdf.Gardella_Mateo_IntellectualProperty.pdf.
Gardella_Mateo_IntellectualProperty.pdf.
MateoGardella
 
Making and Justifying Mathematical Decisions.pdf
Making and Justifying Mathematical Decisions.pdfMaking and Justifying Mathematical Decisions.pdf
Making and Justifying Mathematical Decisions.pdf
Chris Hunter
 

Último (20)

Mixin Classes in Odoo 17 How to Extend Models Using Mixin Classes
Mixin Classes in Odoo 17  How to Extend Models Using Mixin ClassesMixin Classes in Odoo 17  How to Extend Models Using Mixin Classes
Mixin Classes in Odoo 17 How to Extend Models Using Mixin Classes
 
An Overview of Mutual Funds Bcom Project.pdf
An Overview of Mutual Funds Bcom Project.pdfAn Overview of Mutual Funds Bcom Project.pdf
An Overview of Mutual Funds Bcom Project.pdf
 
1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdf
 
Nutritional Needs Presentation - HLTH 104
Nutritional Needs Presentation - HLTH 104Nutritional Needs Presentation - HLTH 104
Nutritional Needs Presentation - HLTH 104
 
fourth grading exam for kindergarten in writing
fourth grading exam for kindergarten in writingfourth grading exam for kindergarten in writing
fourth grading exam for kindergarten in writing
 
Unit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptxUnit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptx
 
Seal of Good Local Governance (SGLG) 2024Final.pptx
Seal of Good Local Governance (SGLG) 2024Final.pptxSeal of Good Local Governance (SGLG) 2024Final.pptx
Seal of Good Local Governance (SGLG) 2024Final.pptx
 
Web & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfWeb & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdf
 
Unit-IV; Professional Sales Representative (PSR).pptx
Unit-IV; Professional Sales Representative (PSR).pptxUnit-IV; Professional Sales Representative (PSR).pptx
Unit-IV; Professional Sales Representative (PSR).pptx
 
Gardella_PRCampaignConclusion Pitch Letter
Gardella_PRCampaignConclusion Pitch LetterGardella_PRCampaignConclusion Pitch Letter
Gardella_PRCampaignConclusion Pitch Letter
 
This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
 
Gardella_Mateo_IntellectualProperty.pdf.
Gardella_Mateo_IntellectualProperty.pdf.Gardella_Mateo_IntellectualProperty.pdf.
Gardella_Mateo_IntellectualProperty.pdf.
 
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
 
Measures of Central Tendency: Mean, Median and Mode
Measures of Central Tendency: Mean, Median and ModeMeasures of Central Tendency: Mean, Median and Mode
Measures of Central Tendency: Mean, Median and Mode
 
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
 
Making and Justifying Mathematical Decisions.pdf
Making and Justifying Mathematical Decisions.pdfMaking and Justifying Mathematical Decisions.pdf
Making and Justifying Mathematical Decisions.pdf
 
psychiatric nursing HISTORY COLLECTION .docx
psychiatric  nursing HISTORY  COLLECTION  .docxpsychiatric  nursing HISTORY  COLLECTION  .docx
psychiatric nursing HISTORY COLLECTION .docx
 
ICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptxICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptx
 
Holdier Curriculum Vitae (April 2024).pdf
Holdier Curriculum Vitae (April 2024).pdfHoldier Curriculum Vitae (April 2024).pdf
Holdier Curriculum Vitae (April 2024).pdf
 

Reconsidering baron and kenny

  • 1. Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis Xinshu Zhao, John G. Lynch Jr., and Qimei Chen (JCR, 2010) Minhwan Lee 2013 (Best Article for 2010), JCR Best Article Award Winners
  • 2. Introduction  Misapplication of the Baron and Kenny  Data do not comport with one or more of the Baron and Kenny criteria  1) cause authors to drop projects that may be promising  2) cause journals to reject paper that may deserve publication  Misunderstanding of mediation: Ignore important hints for theory building  Baron and Kenny (1986): cited by 44,155, June 2014, Google scholar  While the popularity of the Baron-Kenny procedure continues to grow, a small technical literature has grown alongside showing flaws in Baron and Kenny’s logic  Fitzsimons (2008) and Irwin and McClelland (2001) translated an existing technical literature on moderated regression for practicing consumer researchers  Similarly, we aim to explain to users of Baron and Kenny’s tests how new developments should change how they test for mediation -2-
  • 3. Baron and Kenny’s Tests  1) Baron and Kenny claim that mediation is strongest when there is an indirect effect but no direct effect in equation 3  2) there need not be a significant “effect to be mediated” in equation 2.  3) the Sobel test is low in power compared to a bootstrap test popularized by Preacher and Hayes (2004), in some cases markedly so. -3-
  • 4. Mediators Hidden in “Direct” Effects: Boon to Theory Building  Baron and Kenny (1986)  “full mediation”: indirect effect but no direct effect  “partial mediation”: both indirect and direct effects  Iacobucci (2008, 12): “when all tests are properly conducted and reported, the majority of articles conclude with ‘partial mediation.”  mediation is usually accompanied by a direct effect  The concept of a “direct” effect is clear statistically, but it is often unclear theoretically  Bolton, Cohen, and Bloom’s (2006), Mitra and Lynch (1995)  “direct” paths often result from omission of one or more mediators from the model (Shrout and Bolger 2002):  there is a silver lining in “partial mediation.”: The sign of the mysterious “direct” effect has heuristic value for theory building. -4-
  • 5. No Need for an “Effect to be mediated”  Competitive and complementary mediations are equally likely and of equal theoretical interest a priori  c and a×b are of the same sign: complementary mediation  c and a×b are of the opposite sign: competitive mediation  only complementary mediations should be judged to be publishable  Hypothetical modification of Mitra and Lynch’s (1995)  the unexplained negative direct path to deter publishing findings of a positive indirect path  (a) advertising increases consideration set size, which in turn increases price sensitivity; (b) despite the progress in the paper, there is room for future work accounting for the direct effect; and (c) in searching for added mediators, future authors should focus first on those that would produce a negative indirect path  Focused discussion would be considerably more helpful than most “Future Research” sections now published -5-
  • 6. A Typology of Mediations and Nonmediations  three patterns consistent with mediation and two with nonmediation  complementary mediation: “consistent” model or “positive confounding”  competitive mediation: “inconsistent” model or “negative confounding”  Proper interpretation of one’s data requires two dimensions for the indirect path and the direct path -6- Five Typology (Suggested) Baron and Kenny 1 Complementary mediation: Mediated effect (a×b) and direct effect (c) both exist and point at the same direction overlaps with partial mediation 2 Competitive mediation: Mediated effect (a×b) and direct effect (c) both exist and point in opposite directions no mediation 3 Indirect-only mediation: Mediated effect (a×b) exists, but no direct effect overlaps with full mediation 4 Direct-only nonmediation: Direct effect (c) exists, but no indirect effect no mediation 5 No-effect nonmediation: Neither direct effect nor indirect effect exists no mediation
  • 7. A Typology of Mediations and Nonmediations  In our approach to mediation analysis, c’ now represents only the total effect— not the “effect to be mediated.”  The X-Y test is never relevant to establishing mediation  be possible to establish an indirect effect despite no total effect  Figure 2*: Decision Tree for Establishing and Understanding Types of Mediation and Nonmediation  decision tree to conceptualize these five types of mediation and nonmediation to convey to readers what really matters in a mediation analysis  The top of the figure (2a) shows the statistical path to establishing mediation and classifying its type  The bottom of the figure (2b) shows the interpretation of the data pattern for conclusions about theory -7-
  • 8. A Typology of Mediations and Nonmediations -8- Baron and Kenny’s three equations are useful; Sobel test, by a superior bootstrap test (Preacher and Hayes 2004, 2008) The main role for Baron and Kenny’s three equations is in deciding the type of mediation
  • 9. A Typology of Mediations and Nonmediations -9- all three cases on the left, the data support the hypothesized mediation story X→M→Y For both complementary and competitive mediations, the significant direct effect c points to the possible existence of some omitted second mediator that can be pursued in future research the sign of the direct effect can point to as yet undiscovered mediators
  • 10. Sobel’s Not Noble  Preacher and Hayes (2004) made a strong case that it is insufficient to show that the effect of X on Y is reduced in size when M is added to the model  the main contribution of Preacher and Hayes (2004) was to present SAS and SPSS syntax for an alternative “bootstrap” test of the indirect effect that is almost always more powerful than Sobel’s test  95% confidence interval will often improperly include zero, compared to the 95% confidence interval we would create if we could observe the sampling distribution of a×b  Zhao (1997): collected during Super Bowl 1994 to show how to perform Preacher-Hayes bootstrap tests in either SAS or SPSS  http://www.comm.hkbu.edu.hk/zhao/shared -10-
  • 11. Sobel’s Not Noble  To perform bootstrapping in SPSS, open your data set and then follow the steps below:  i. Open the Preacher-Hayes script in SPSS for Windows (.sbs file).  ii. Run the Preacher-Hayes script from the scripting window to activate the dialog box.  iii. Identify your dependent (i.e., liking), independent (frequency), and one or more mediating variables (here only clutter). You may also identify covariate variables if appropriate.  iv. Set bootstrap samples to, say, 5,000.  v. Set confidence level at 95% or 99%.  vi. Click OK to execute the script.  vii. Find “Bootstrap Results for Indirect Effects” from the output window and the 95% confidence interval (-.0930 to -.0268 in our case). If the confidence interval does not include 0, the indirect effect a×b is significant and mediation is established, which takes you to the left of figure 2. If the confidence interval includes 0, a×b is not significant and mediation hypothesis is rejected, which takes you to the right of figure 2. -11-
  • 12. Sobel’s Not Noble -12-  Open data: 1994 Super Bowl Data SPSS  Analyze- Regression - Preacher and Hayes (2008) Multiple Mediation (INDIRECT)
  • 13. Sobel’s Not Noble -13-  Results (Consistent with Figure 3)
  • 14. Sobel’s Not Noble -14-  Results (Consistent with Figure 3)
  • 15. Recommended Steps for Mediation Analysis  the indirect effect is significant  Simply run the Preacher Hayes script and generate “Bootstrap Results for Indirect Effects,” as we showed in figure 3, to determine whether the indirect effect a×b is significant and thus whether to take the left (mediation) or right (nonmediation) branch of figure 2  Use SEM to estimate all parameters simultaneously, or you may run the two regression equations 1 and 3; Preacher-Hayes script output (fig. 3) provides these estimates automatically under “Direct and Total Effects.”  i. If a×b is significant but c is not, you have indirect only mediation.  ii. If a×b is not significant but c is, you have direct only nonmediation.  iii. If neither a×b nor c is significant, you have no effect nonmediation.  iv. If both a×b and c are significant, determine the sign of a×b×c by multiplying the three coefficients, or by multiplying c by the mean value of a×b from the bootstrap output. If a×b×c is positive, it is complementary mediation; if a×b×c is negative, it is competitive mediation. -15-
  • 16. Recommended Steps for Mediation Analysis  Report also the unstandardized regression coefficients a, b, and c to allow substantive interpretation of the results -16- • a×b=-.0527 • 95% confidence interval excluding zero (-.0930 to -.0268) • a=2.53 • b=-.0208 • c=.057 • a×b×c=-.0030; negative/competitive mediation
  • 17. Unexpectedly Flipped Signs of Indirect Paths  Cohen and Cohen (1975) and Friedman and Wall (2005) note that an indirect effect opposite in sign to the zero-order correlations arises in cases of “net suppression”  All cases of net suppression in mediation analysis can be seen as special cases of competitive mediation in which the direct path dominates the indirect  To avoid publishing a conclusion that is wrong, authors should report (and be asked by journals to report) the type of mediation and actual a, b, and c coefficients, not just the significance test of the indirect effect -17-
  • 18. Structural Equation Models and Baron and Kenny’s Regression Approach  Iacobucci (2008) argues that structural equation (SEM) approaches dominate the “causal steps” approach of Baron and Kenny (1986)  SEM approach is superior to Baron and Kenny’s because it estimates everything simultaneously instead of assuming that equations 1–3 are independent  1) even if there is no “true” direct effect or omitted second mediator, one can observe a significant direct effect due to measurement error in the indicator of M(Birnbaum and Mellers 1979)  2) sometimes a researcher’s data may seem to conform to Baron and Kenny’s conditions for mediation, but the “mediator” is not conceptually different from the independent variable: it is effectively a manipulation check -18-
  • 19. Conclusion  We have simplified these criticisms and added our own, providing an overarching framework that considers two dimensions—the indirect effect and the direct effect—rather than the one dimensional “full,” “partial,” and “none” classification employed by Baron and Kenny.  1) although Baron and Kenny and current practice hold up “full mediation” as the gold standard, most studies report “partial mediation” with a significant direct path c  2) there need not be a significant rXY in a proper mediation analysis; The only requirement for mediation is that the indirect effect a×b be significant  3) in testing the significance of an indirect effect a×b, use the more rigorous and powerful bootstrap test, not Sobel -19-

Notas do Editor

  1. 많은 연구에서 Baron and Kenny의 Mediation을 사용하지만 종종 Data의 사용 법에 있어서 적절하지 않게 사용하는 경우가 있음 이러한 Misapplication은 연구자가 설정한 가설 증명을 어렵게 하기도 하며, publish 할 수 있는 페이퍼가 Reject 당하는 경우를 만들어내기도 함 따라서 Mediation을 잘못 이해하는 것은 이론 수립에 있어서 중요한 힌트를 무시하는 것과 같은 결과를 도출할 수 있음 Baron and Kenny의 방법은 많은 연구에서 활용되며 큰 인기를 얻었으나, 기술적인 부분에 대한 발전은 이루어지지 않았음 몇몇의 연구자들은 이러한 방법론의 발전에 대하여 주장하였고, 본 연구는 비슷하게 Mediation을 활용하는 데에 있어서 어떠한 방법을 사용해야 하는지에 대한 가이드라인을 제공하는 것을 목적으로 함
  2. 기존의 Baron and Kenny의 방법론은 좌측의 Equation (1)~(3)까지의 수식을 바탕으로 mediated effect를 보는 것에 초점을 맞추고 있음 a는 독립변수가 매개변수에, b는 매개변수가 종속변수에, c는 독립변수가 종속변수에 미치는 영향의 선을 나타내는 계수이며, c’는 a와 b가 통제되었을 때의 독립변수가 매개변수를 통하여 종속변수에 미치는 영향을 나타내는 계수 Equation (4)는 Preacher and Hayes (2004)에서 주장된 Bootstrap test의 식을 나타내고, 이를 일반적으로 Sobel z-test라고 명명함 1) Baron and Kenny는 mediation이 강할 때 equation (3)의 direct effect는 없다라고 주장하였지만, mediation의 strength는 indirect effect의 크기에 의해서 측정되기 때문에 direct effect의 존재는 다른 mediator에 대한 theorizing에 대하여 알리는 역할을 수행 2) equation (2)의 mediated effect는 유의할 필요성이 없는데, mediation이 수립되기 위해서는 a와 b의 indirect effect가 유의하게 나타나면 됨; 또한 Baron and Kenny의 방법은 주로 이러한 mediation을 분류하는 데에 주요한 역할을 수행함 3) Sobel test는 bootstrap test에 비해서 power가 약하고, 어떠한 경우에는 이것이 크게 나타나기도 함; 때때로 a와 b의 positive indirect effect를 기대하는 연구자들은 X와 Y, X와 M, Y와 M의 관계의 방향성을 무시하는 경우가 발생함
  3. Baron and Kenny는 direct effect가 없고 indirect effect만 존재하는 것을 full mediation이라고 지칭했고, direct, indirect effect가 모두 존재하는 것을 partial mediation이라고 지칭하였으나, Iacobucci(2008)은 대부분의 연구가 partial mediation으로 결론을 짓고 있기 때문에 mediation은 대부분 direct effect와 함께 존재한다라는 사실을 제시함 이처럼 direct effect는 통계적으로는 명확하지만 이론적으로 불명확한 경우가 많음 예를 들어 Bolton, Cohen, and Bloom(2006)의 연구를 figure 1에 적용해서 살펴보면, path a는 콘돔의 사용가능성은 인지된 위험을 낮춰주고, path b는 인지된 위험은 질병의 가능성을 증가시킨다는 결과를 제시하였음; 하지만 명확하게 콘돔의 사용가능성은 콘돔의 사용으로 인한 신체적인 보호로 질병의 발병을 감소시켜주지만, 결론적으로는 negative direct path c를 형성하게 됨  mediated path a와 b는 positive이지만, 인지된 위험이 콘돔의 이용가능성을 완벽하게 매개하지 못한다면 get-out-of-jail-free cards 이론을 약화시키지 않을 수 있음 다른 예로는 Mitra and Lynch (1995)는 2개의 mediator를 통하여 advertising이 price sensitivity에 미치는 영향에 대하여 실험적으로 살펴보았는데, (1) consideration set size를 증가시켜서 이것이 가격탄력성을 증가시킬 수 있고, (2) 경쟁제품간의 인지된 효용간의 차이를 증가시켜서 가격탄력성을 감소시킬 수 있는데, 이 경우 연구자가 두 개의 mediator에서 첫 번째 mediator만을 사용하게 된다면(advertising  consideration set size  price sensitivity) indirect path axb는 positive로 나타날 수 있지만, 기대되지 않았고 정해지지 않은 direct path c는 negative일 수 있음; 이 경우에 연구자는 기대되지 않은 negative direct effect를 보고하고 추가적인 선행연구 리뷰를 통해서 다른 연구자들로 하여금 두 번째 mediator를 찾게끔 후속 연구를 독려할 수 있음 뿐만 아니라 relationship marketing 에서의 다양한 예시가 존재하기도 함 이처럼 direct path는 종종 수립된 모델로부터 가능한 더 많은 mediator의 누락하게 되는 결론에 도달할 수 있음; 따라서 partial mediation에 의해서 direct effect의 방향성의 애매모호성에 의거하여 theory building에 있어서 heuristic value를 가지기 때문에 연구자의 자의적인 해석이 난무할 수 있음
  4. 앞서 제시된 Baron and Kenny의 Equation들을 보면 Equation (2)는 mediated effect를 보여주고 있는데, 이러한 효과를 검증하는 것이 굳이 필요하지 않고, direct effect의 방향성에 의거하여 mediation effect를 검증하는 것이 가능 Competitive, complementary는 같은 이론적 배경을 바탕으로 수립될 수 있는데 c와 axb의 방향성이 같으면 complementary, 다르면 competitive mediation이라고 볼 수 있지만, academic field에서는 일반적으로 complementary mediation만이 좋은 연구결과로 취급을 받음 앞서 말했던 Mitra and Lynch’s (1995) 연구의 광고가 가격탄력성에 미치는 영향을 살펴보면, 기대되지 않은 negative direct path는 positive indirect path에 대한 finding을 게재하는 것을 방해할 수 있고, 이 때 연구의 제한 점으로 (1) 광고는 consideration set size를 증가시키고, 가격 탄력성을 증가시키지만, (2) direct effect를 설명할 수 있는 후속 연구가 필요하고, (3) 추가할 수 있는 mediator를 찾아야 하고 후속 연구자는 해당 연구의 negative indirect path를 만들어내는 것에 주목해야 한다고 설명할 수 있을 것  이처럼 단순히 mediation만을 제공하는 것이 아니라 direct, indirect의 효과에 대한 방향성을 제공하는 것이 이론적인 발전에 큰 도움을 줄 수 있음
  5. 앞선 연구결과를 바탕으로 3가지 패턴의 mediation과 2가지의 nonmediation을 아래와 같이 정리할 수 있음 1) complementary mediation은 axb와 direct c가 존재하고 같은 방향일 때를 의미하며, 이것은 Baron and Kenny의 partial mediation과 동일하다고 볼 수 있음 2) competitive mediation은 axb와 direct c가 존재하나 다른 방향일 때를 의미하며, 이것은 Baron and Kenny의 nonmediation임 3) indirect-only mediation은 axb는 존재하지만 direct c는 존재하지 않는 때를 의하며, 이것은 Baron and Kenny의 full mediation과 동일하다고 볼 수 있음 4) direct-only nonmediation은 axb의 indirect는 없고, direct c만 존재하는 것을 의미하며, 이것은 Baron and Kenny의 nonmediation임 5) no-effect nonmediation은 direct, indirect는 모두 존재하지 않는 것을 의미하며, 이것은 Baron and Kenny의 nonmediation임 Complementary mediation은 consistent model, positive confounding이라고 볼 수 있고, 반대로 competitive mediation은 inconsistent model, negative confounding으로 볼 수 있음 따라서 mediation에 대한 적절한 해석은 indirect와 direct effect에 대한 2가지 차원에 기반한 해석이 필요함
  6. 따라서 본 연구의 목적은 total effect만을 제공할 것이 아니라 각각의 변수간의 관계뿐만 아니라 방향성을 제공해야만 연구결과에 대한보다 풍부한 해석이 가능하고, total effect에 대한 고려 없이 direct, indirect effect 만으로도 mediation에 대한 해석이 가능할 수 있음 본 연구에서 말하고자 하는 내용은 mediation, nonmediation에 따른 유형을 수립하고 형성하는데 위한 의사결정 나무를 표현한 Figure 2에 모든 내용이 담겨져 있음; 상반부인 2a는 mediation의 유형을 구분 짓고, mediation을 형성하는 path에 대한 통계적인 부분에 대한 것이고, 하반부인 2b는 이론에 대한 결과의 데이터 패턴에 대한 해석을 보여주는 것
  7. Baron and Kenny의 3가지의 equation도 유용하지만, sobel test와 bootstrap이 훨씬 유용한 도구 Baron and Kenny의 criteria는 mediation의 유형을 결정짓는데 영향을 미침
  8. Mediation에 대한 부분은 모두 X-M-Y의 관계를 지지하는 데이터로 볼 수 있고, complementary, competitive mediation 모두에서의 유의미한 direct effect c는 향후 연구를 위하여 mediator를 찾기 위해서 2번째의 누락된 mediator 찾는 것을 가능하게 해줌 Nonmediation은 유의미한 결과를 제공하지는 못하지만, direct effect의 방향은 아직 발견되지 않은 mediator의 존재를 알려줄 수 있음
  9. Preacher and Hayes (2004)는 SPSS와 SAS를 활용하여 bootstrap을 활용할 수 있게 만든 업적을 쌓았고, 이와 관련된 syntax, macro를 별도로 제공하기 때문에 쉽게 sobel test, indirect test를 쉽게 할 수 있음 본 논문에서 소개된 Zhao (1997)은 슈퍼볼 기간 동안에 브랜드 별 광고 데이터를 바탕으로 어떻게 bootstrap을 활용할 수 있는지에 대한 데이터와 결과, 명령문 등을 홈페이지에서 제공하고 있음
  10. 본 논문에서 제시된 bootstrapping을 SPSS에서 실행시키는 step은 다음과 같음; step by step의 이미지를 보고 설명하겠음
  11. 먼저 홈페이지에서 제공하는 데이터를 다운받아서 실행시키고, 미리 깔아놓은 indirect; Preacher and Hayes (2008)을 실행시키고, 종속변수, 매개변수, 독립변수를 넣어주고, Sample의 수는 5000, 신뢰수준은 95%로 설정하여 OK를 눌러주면 끝!
  12. 위를 수행하면 아래와 같은 결과가 나옴; 결과에 대한 해석은 뒷부분에서 계속하도록 하겠음
  13. 실행한 결과는 Figure 3과 동일
  14. Mediation analysis를 위해서는 결과에 대한 해석을 제공할 수 있는 a, b, c에 대한 unstandardized regression coefficient를 레포트 해 주어야 함 해당 결과를 살펴보면 먼저 axb의 indirect effect는 -.0527이고, 신뢰구간은 -.0930 에서 -.0268로 0을 포함하지 않기 때문에 mediation이 유의미하다고 볼 수 있음 분석결과에 따른 각각의 회귀계수를 보면 a=2.53, b=-.0208, c=.057이고 이를 모두 곱한 a×b×c=-.0030로 negative, 즉 음수의 값을 갖기 때문에 competitive mediation이라고 볼 수 있음
  15. 몇몇의 연구자들은 indirect effect가 반대되는 방향성을 가지는 것은 계수간의 correlation으로 net suppression때문에 발생할 수 있지만, 이는 흔한 경우는 아님 잘못된 결과를 보고하는 것을 막기 위해서는 indirect effect의 유의성만 보고하는 것이 아니라 각각의 계수에 대한 회귀계수를 보고하는 것이 바람직함
  16. Iacobucci (2008)는 SEM을 활용하면 Baron and Kenny의 방법보다 인과적인 단계를 바탕으로 실행할 수 있다고 했는데, Baron and Kenny가 독립적으로 Equation 1~3까지를 가정하는 것 대신에 SEM은 동시에 모든 것을 추정할 수 있기 때문에 보다 좋은 방법일 수 있음 하지만 SEM을 사용할 때에도 아래와 같은 2가지 사항에 대한 주의가 필요함 1) true direct effect 나 누락된 second mediator가 없다고 하더라도 매개변수의 indicator에서의 발생할 수 있는 measurement error때문에 유의미한 direct effect c가 발생할 수 있음; multiple measure로 측정되는 매개변수는, mediator의 측정에서의 발생하는 SEM Model error가 매개변수를 측정하는 것에 있어서 인위적인 error를 제거해주고, true direct effect를 측정하게 할 수 있음 2) 대부분의 연구자들이 Baron and Kenny의 mediation condition에 동의하지만 mediator와 independent variable은 개념상으로 다르기 때문에 manipulation check이 효과적임
  17. 본 연구는 mediation analysis에 대한 통합적인 프레임워크를 제공하고자 했으며, Baron and Kenny의 단순한 full, partial mediation으로 구분 짓는 것에서 벗어나 direct, indirect effect의 두 가지 관점에서 mediation analysis를 하는 것이 바람직하다고 주장하고 있음 본 연구의 주된 내용을 짧게 다시 한번 요약하면, 1) Baron and Kenny는 full mediation이 황금률인데도 불구하고 대부분의 연구들은 partial mediation 결과를 나타내고 있고, direct path는 거의 설명되거나 예측되지 않음; 설명되지 않은 direct path는 omitted mediator를 가리킬 수 있기 때문에, 이를 향후 연구에 제시하는 것이 바람직함 2) Mediation analysis에서 direct c의 유의미성은 꼭 필요하지 않음; 중요한 것은 axb의 indirect effect의 유의성임 3) Mediation analysis에서 가장 중요한 indirect axb를 분석하기 위해서 sobel test보다 엄격하고 강력한 bootstrap test를 사용하는 것이 좋음