bayes factors as a replacement for t-tests
TRANSCRIPT
Bayes factors as a replacement for t-tests
Jeffrey N. Rouder
University of Missouri
September, 2008
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Collaborators
I Paul Speckman
I Dongchu Sun
I Richard Morey
I Geoff Iverson
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
The De Facto Rule of Data Analysis
p < .05 −→ Good
p ≥ .05 −→ Bad
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
The De Facto Rule of Data Analysis
p < .05 −→ Good
p ≥ .05 −→ Bad
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Mission of This Talk
Provide a practical and useful alternative to the De Facto Rule inData Analysis.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Plan
1. Invariances, as opposed to differences, are the heart of science
2. Significance tests are ill-suited for assessing invariances
3. Significance tests are ill-suited for assessing differences
4. Bayes factor approach
5. Easy-to-use web applet to speed adoption
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Johannes Kepler (1571-1630)
I Planets varied greatly inthe speed & direction oftheir paths through thesky.
I Kepler extracted theinvariants of celestialmotion (e.g., ellipticalorbits, equal areacircumscribed in equaltime).
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Johannes Kepler (1571-1630)
I Planets varied greatly inthe speed & direction oftheir paths through thesky.
I Kepler extracted theinvariants of celestialmotion (e.g., ellipticalorbits, equal areacircumscribed in equaltime).
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Johannes Kepler (1571-1630)
I Planets varied greatly inthe speed & direction oftheir paths through thesky.
I Kepler extracted theinvariants of celestialmotion (e.g., ellipticalorbits, equal areacircumscribed in equaltime).
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
1. Usually, observables change across conditions.
2. What relations among observables remain constant orinvariant?
3. These invariances form phenomena to be explained by theory.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
1. Usually, observables change across conditions.
2. What relations among observables remain constant orinvariant?
3. These invariances form phenomena to be explained by theory.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
1. Usually, observables change across conditions.
2. What relations among observables remain constant orinvariant?
3. These invariances form phenomena to be explained by theory.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
I Conservation Laws: e.g., F = MA implies that if two objectsare dropped from the same height F1
M1= F2
M2. Moreover,
conservation laws hold nearly exactly across everydaystiuations.
I In genetics, adenine binds to thymine, guanine binds tocytosine, across all DNA in all species.
I In chemistry, mecahisms of covalent bonding are the sameacross all atoms.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
I Conservation Laws: e.g., F = MA implies that if two objectsare dropped from the same height F1
M1= F2
M2. Moreover,
conservation laws hold nearly exactly across everydaystiuations.
I In genetics, adenine binds to thymine, guanine binds tocytosine, across all DNA in all species.
I In chemistry, mecahisms of covalent bonding are the sameacross all atoms.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
I Conservation Laws: e.g., F = MA implies that if two objectsare dropped from the same height F1
M1= F2
M2. Moreover,
conservation laws hold nearly exactly across everydaystiuations.
I In genetics, adenine binds to thymine, guanine binds tocytosine, across all DNA in all species.
I In chemistry, mecahisms of covalent bonding are the sameacross all atoms.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances At The Heart of Science
I Conservation Laws: e.g., F = MA implies that if two objectsare dropped from the same height F1
M1= F2
M2. Moreover,
conservation laws hold nearly exactly across everydaystiuations.
I In genetics, adenine binds to thymine, guanine binds tocytosine, across all DNA in all species.
I In chemistry, mecahisms of covalent bonding are the sameacross all atoms.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances in Psychology?
p < .05 −→ Good
p ≥ .05 −→ Bad
Effects are valued; lack of effects are not.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances in Psychology?
p < .05 −→ Good
p ≥ .05 −→ Bad
Effects are valued; lack of effects are not.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances in Psychology?
p < .05 −→ Good
p ≥ .05 −→ Bad
Effects are valued; lack of effects are not.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Gender (Shibley Hyde, 2005, 2007)
I Performance on many tasks is invariant to gender
I How can these invariances come about?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Weber’s Law (1860)
I ∆ ∝ I
I For two different backgrounds, ∆1I1
= ∆I2I2
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Choice Rule (Clarke, 1957; Luce, 1959; Shepard, 1957)
I Choice probabilities:
I Gin & Tonic .5I Beer .4I Wine .1
I Out of gin
I Beer .8I Wine .2
I Invariance in the ratio of choice probabilities (4:1)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Choice Rule (Clarke, 1957; Luce, 1959; Shepard, 1957)I Choice probabilities:
I Gin & Tonic .5I Beer .4I Wine .1
I Out of gin
I Beer .8I Wine .2
I Invariance in the ratio of choice probabilities (4:1)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Choice Rule (Clarke, 1957; Luce, 1959; Shepard, 1957)I Choice probabilities:
I Gin & Tonic .5I Beer .4I Wine .1
I Out of ginI Beer .8I Wine .2
I Invariance in the ratio of choice probabilities (4:1)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Choice Rule (Clarke, 1957; Luce, 1959; Shepard, 1957)I Choice probabilities:
I Gin & Tonic .5I Beer .4I Wine .1
I Out of ginI Beer .8I Wine .2
I Invariance in the ratio of choice probabilities (4:1)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Law-like Phenomena: Psychometric Functions Shift (Watson &Pelli, 1983)
Intensity
Pro
babi
lity
0.5
0.7
0.9
10 50 100 500 1000 5000
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Cowan’s K model (2001)
I H = Pr(“change” | change)
I F = Pr(“change” | same)
IH1 − F1
H2 − F2=
N2
N1, N1,N2 > K
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Cowan’s K model (2001)
I H = Pr(“change” | change)
I F = Pr(“change” | same)
IH1 − F1
H2 − F2=
N2
N1, N1,N2 > K
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Cowan’s K model (2001)
I H = Pr(“change” | change)
I F = Pr(“change” | same)
IH1 − F1
H2 − F2=
N2
N1, N1,N2 > K
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Examples of Invariances in Psychology
Selective Influence, Sternberg (1969)
I For a given model, a manipulation should affect someparameters and not others.
I Example: Process Dissociation. Dividing attention at testshould affect recollection but not automatic activation.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 1: Double Dissociations
Conscious Familiar Conscious Familiar
LowHigh
Val
ue
0.0
0.2
0.4
0.6
0.8
1.0
Divided Attention Percpetual Similarity
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 1: Double Invariance
Conscious Familiar Conscious Familiar
LowHigh
Val
ue
0.0
0.2
0.4
0.6
0.8
1.0
Divided Attention Percpetual Similarity
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.
I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.
I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.
I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Note 2: Maybe there are no invariances
I Antinull View; championed by Meehl, Cohen
I Examples: Newtonian mechaniscs.I Constructive Challenge:
I Invariances may exist at a platonic level but be perturbed inthe real world.
I The goal is to find platonical rather than actual invariances.I Anticipates subjectivity in inference.I Pragmatics: Emphasis on model selection rather than truth
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances Are Difficult To Assess
I Invariances are null hypotheses
I Significance tests cannot provide evidence for the null.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances Are Difficult To Assess
I Invariances are null hypotheses
I Significance tests cannot provide evidence for the null.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Invariances Are Difficult To Assess
I Invariances are null hypotheses
I Significance tests cannot provide evidence for the null.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Problem: One-Sample Design
I Is there a difference?
I Difference Scores: y1, y2, . . . , yN
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I Test: Paired t-test. Calculate p. Is p < .05?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Problem: One-Sample Design
I Is there a difference?
I Difference Scores: y1, y2, . . . , yN
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I Test: Paired t-test. Calculate p. Is p < .05?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Problem: One-Sample Design
I Is there a difference?
I Difference Scores: y1, y2, . . . , yN
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I Test: Paired t-test. Calculate p. Is p < .05?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Problem: One-Sample Design
I Is there a difference?
I Difference Scores: y1, y2, . . . , yN
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I Test: Paired t-test. Calculate p. Is p < .05?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Problem: One-Sample Design
I Is there a difference?
I Difference Scores: y1, y2, . . . , yN
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I Test: Paired t-test. Calculate p. Is p < .05?
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
The t-test
−4 −2 0 2 4 6 8
0.0
0.1
0.2
0.3
0.4
t−value
Den
sity
H0
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
The t-test
−4 −2 0 2 4 6 8
0.0
0.1
0.2
0.3
0.4
t−value
Den
sity
H0
H1
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Significance Tests
−3 −1 1 3
0.0
0.1
0.2
0.3
0.4
Difference
Den
sity
−4 0 4 8
0.0
0.1
0.2
0.3
t−value
Den
sity
N == 10
0.0 0.4 0.8
02
46
810
p−value
Den
sity
N == 10
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Significance Tests
−3 −1 1 3
0.0
0.1
0.2
0.3
0.4
Difference
Den
sity
−4 0 4 8
0.0
0.1
0.2
0.3
t−value
Den
sity
N == 100
0.0 0.4 0.8
02
46
810
12
p−value
Den
sity
N == 100
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Significance Tests
−3 −1 1 3
0.0
0.1
0.2
0.3
0.4
Difference
Den
sity
−4 0 4 8
0.0
0.1
0.2
0.3
0.4
t−value
Den
sity
N == 10
0.0 0.4 0.8
0.0
0.2
0.4
0.6
0.8
1.0
p−value
Den
sity
N == 10
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Significance Tests
−3 −1 1 3
0.0
0.1
0.2
0.3
0.4
Difference
Den
sity
−4 0 4 8
0.0
0.1
0.2
0.3
0.4
t−value
Den
sity
N == 100
0.0 0.4 0.8
0.0
0.2
0.4
0.6
0.8
1.0
p−value
Den
sity
N == 100
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Fantasy J-Value Statistic
0.0 0.2 0.4 0.6 0.8 1.0
05
1015
20
J−value
Den
sity
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Resulting Critiques of Significance Tests
I Consequence #1: Can’t gain evidence for the null.
I Consequence #2: Overstates the evidence against the null.
I Berger & Berry (1988): Miscalibration
I Meehl (1978): Design flaw from the faulty reasoning ofPopper and Fisher
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Resulting Critiques of Significance Tests
I Consequence #1: Can’t gain evidence for the null.
I Consequence #2: Overstates the evidence against the null.
I Berger & Berry (1988): Miscalibration
I Meehl (1978): Design flaw from the faulty reasoning ofPopper and Fisher
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Resulting Critiques of Significance Tests
I Consequence #1: Can’t gain evidence for the null.
I Consequence #2: Overstates the evidence against the null.
I Berger & Berry (1988): Miscalibration
I Meehl (1978): Design flaw from the faulty reasoning ofPopper and Fisher
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bias to Overstate the Evidence Against the Null
I Example: N = 100, y = 10ms, t = 2.38, p ≈ .02.
I Consider the alternative µ = 30ms, which is a typical effect,e.g., priming.
I Which is more likely given the data, the null (µ = 0) or thetypical alternative (µ = 30)?
I Compute Likelihood ratio
L(µ = 0; data)
L(µ = 30; data)≈ 3800
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bias to Overstate the Evidence Against the Null
I Example: N = 100, y = 10ms, t = 2.38, p ≈ .02.
I Consider the alternative µ = 30ms, which is a typical effect,e.g., priming.
I Which is more likely given the data, the null (µ = 0) or thetypical alternative (µ = 30)?
I Compute Likelihood ratio
L(µ = 0; data)
L(µ = 30; data)≈ 3800
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bias to Overstate the Evidence Against the Null
I Example: N = 100, y = 10ms, t = 2.38, p ≈ .02.
I Consider the alternative µ = 30ms, which is a typical effect,e.g., priming.
I Which is more likely given the data, the null (µ = 0) or thetypical alternative (µ = 30)?
I Compute Likelihood ratio
L(µ = 0; data)
L(µ = 30; data)≈ 3800
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bias to Overstate the Evidence Against the Null
I Example: N = 100, y = 10ms, t = 2.38, p ≈ .02.
I Consider the alternative µ = 30ms, which is a typical effect,e.g., priming.
I Which is more likely given the data, the null (µ = 0) or thetypical alternative (µ = 30)?
I Compute Likelihood ratio
L(µ = 0; data)
L(µ = 30; data)≈ 3800
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Alternative (ms)
Like
lihoo
d R
atio
(nu
ll/al
tern
ativ
e)
0 10 20 30 40
0.01
110
1000
1e+
05
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Truths About Testing
1. The following methods overstate the evidence against thenull: p-values, confidence intervals, p-rep, Neyman-Pearson(with fixed α), statistical equivalence, Akaike InformationCriterion (AIC).
2. Princibled inference is only possible with a priori specificationof the alternatives. This fact is true for Bayesians and classichypothesis testing.
3. Hypothesis testing is inherently subjective, but don’t freak outyet.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Truths About Testing
1. The following methods overstate the evidence against thenull: p-values, confidence intervals, p-rep, Neyman-Pearson(with fixed α), statistical equivalence, Akaike InformationCriterion (AIC).
2. Princibled inference is only possible with a priori specificationof the alternatives. This fact is true for Bayesians and classichypothesis testing.
3. Hypothesis testing is inherently subjective, but don’t freak outyet.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Truths About Testing
1. The following methods overstate the evidence against thenull: p-values, confidence intervals, p-rep, Neyman-Pearson(with fixed α), statistical equivalence, Akaike InformationCriterion (AIC).
2. Princibled inference is only possible with a priori specificationof the alternatives. This fact is true for Bayesians and classichypothesis testing.
3. Hypothesis testing is inherently subjective, but don’t freak outyet.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factors to the Rescue
Intellectual Legacy:
I Bayes (1760s): Places probability directly on hypotheses
I Laplace (1810s): Proposes using odds for inference
I Jeffreys (1960): Formalizes Bayes factor; proposesspecification of alternative that we adopt here.
I Zellner & Siow (1980): Reformulates Jeffrey’s work broadlyfor linear models.
I Berger, Bayarri & colleagues (2001-2008): ShowedJeffreys-Zellner-Siow Bayes factors have desirablemathemtatical properties.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factors to the Rescue
Intellectual Legacy:
I Bayes (1760s): Places probability directly on hypotheses
I Laplace (1810s): Proposes using odds for inference
I Jeffreys (1960): Formalizes Bayes factor; proposesspecification of alternative that we adopt here.
I Zellner & Siow (1980): Reformulates Jeffrey’s work broadlyfor linear models.
I Berger, Bayarri & colleagues (2001-2008): ShowedJeffreys-Zellner-Siow Bayes factors have desirablemathemtatical properties.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factors to the Rescue
Intellectual Legacy:
I Bayes (1760s): Places probability directly on hypotheses
I Laplace (1810s): Proposes using odds for inference
I Jeffreys (1960): Formalizes Bayes factor; proposesspecification of alternative that we adopt here.
I Zellner & Siow (1980): Reformulates Jeffrey’s work broadlyfor linear models.
I Berger, Bayarri & colleagues (2001-2008): ShowedJeffreys-Zellner-Siow Bayes factors have desirablemathemtatical properties.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factors to the Rescue
Intellectual Legacy:
I Bayes (1760s): Places probability directly on hypotheses
I Laplace (1810s): Proposes using odds for inference
I Jeffreys (1960): Formalizes Bayes factor; proposesspecification of alternative that we adopt here.
I Zellner & Siow (1980): Reformulates Jeffrey’s work broadlyfor linear models.
I Berger, Bayarri & colleagues (2001-2008): ShowedJeffreys-Zellner-Siow Bayes factors have desirablemathemtatical properties.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factors to the Rescue
Intellectual Legacy:
I Bayes (1760s): Places probability directly on hypotheses
I Laplace (1810s): Proposes using odds for inference
I Jeffreys (1960): Formalizes Bayes factor; proposesspecification of alternative that we adopt here.
I Zellner & Siow (1980): Reformulates Jeffrey’s work broadlyfor linear models.
I Berger, Bayarri & colleagues (2001-2008): ShowedJeffreys-Zellner-Siow Bayes factors have desirablemathemtatical properties.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Theorem
Pr(A | B) =Pr(B|A)Pr(A)
Pr(B)
Pr(H | data) =Pr(data|H)× Pr(H)
Pr(data)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Theorem
Pr(A | B) =Pr(B|A)Pr(A)
Pr(B)
Pr(H | data) =Pr(data|H)× Pr(H)
Pr(data)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factor
Ω =Pr(H0 | data)
Pr(H1 | data)
=p(data|H0)
p(data|H1)× Pr(H0)
Pr(H1)
= B01 ×Pr(H0)
Pr(H1),
where
B01 =p(data|H0)
p(data|H1)=
M0
M1
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factor
Ω =Pr(H0 | data)
Pr(H1 | data)
=p(data|H0)
p(data|H1)× Pr(H0)
Pr(H1)
= B01 ×Pr(H0)
Pr(H1),
where
B01 =p(data|H0)
p(data|H1)=
M0
M1
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Bayes Factor
Ω =Pr(H0 | data)
Pr(H1 | data)
=p(data|H0)
p(data|H1)× Pr(H0)
Pr(H1)
= B01 ×Pr(H0)
Pr(H1),
where
B01 =p(data|H0)
p(data|H1)=
M0
M1
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Example
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ1, σ2), µ1 6= 0
I Researcher specifies µ1 before hand.
I Lets also assume σ2 is known.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Example
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ1, σ2), µ1 6= 0
I Researcher specifies µ1 before hand.
I Lets also assume σ2 is known.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Example
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ1, σ2), µ1 6= 0
I Researcher specifies µ1 before hand.
I Lets also assume σ2 is known.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
−10 0 10 20 30 40 50
0.00
00.
005
0.01
00.
015
Effect µµ (ms)
Den
sity
Null Alternative: µµ1 == 40
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Example
M0 = L(y;µ = 0) =1√2πσ
exp
(−
∑y2i
2σ2
)M1 = L(y;µ = µ1)
1√2πσ
exp
(−
∑(yi − µ1)
2
2σ2
)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Simplest Example
Alternative µµ1 (units of σσ)
Bay
es F
acto
r B
01
0.0 0.2 0.4 0.6 0.8 1.0
0.00
10.
110
1000
1e+
05
y == 0
y == 0.2σσ
y == 0.35σσ
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Next Example
Let’s assume σ2 is known.
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I µ ∼ Normal(0, σ20),
I Researcher needs to specify σ20
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Next Example
Let’s assume σ2 is known.
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I µ ∼ Normal(0, σ20),
I Researcher needs to specify σ20
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Next Example
Let’s assume σ2 is known.
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I µ ∼ Normal(0, σ20),
I Researcher needs to specify σ20
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Next Example
Let’s assume σ2 is known.
I Null: yi ∼ Normal(0, σ2)
I Alternative: yi ∼ Normal(µ, σ2), µ 6= 0
I µ ∼ Normal(0, σ20),
I Researcher needs to specify σ20
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
−50 0 50
0.00
00.
005
0.01
00.
015
Effect µµ (ms)
Den
sity
Null
Alternativeσσ0 == 30
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Next Example
M0 = L(y;µ = 0)
M1 =
∫µ
L(y;µ)p(µ)dµ
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Prior Standard Deviation σσµµ (units of σσ)
Bay
es F
acto
r B
01
0.01 0.1 1 10 100 1000 1e+05
1e−
050.
011
100
1e+
05
y == 0
y == 0.15σσ
y == 0.22σσ
B.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Lessons
I As the alternative include a greater percentage ofunrealistically large values, the Bayes factor favors the null.
I If the alternative can include a wide-range of values, it ispenalized by Bayes factor.
I Penalty for complexity
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
What is the Right Value for Variance σ20
I If σ2 is small, then σ20 should be small (perception)
I If σ2 is large, then σ20 should be large (clinical applications)
I Idea: Place prior on effect size instead: δ = µ/σ.
yi ∼ Normal(µ, σ2) ∼ Normal(δσ, σ2).
I δ ∼ Normal(0, σ2δ )
I Default: σ2δ = 1.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
What is the Right Value for Variance σ20
I If σ2 is small, then σ20 should be small (perception)
I If σ2 is large, then σ20 should be large (clinical applications)
I Idea: Place prior on effect size instead: δ = µ/σ.
yi ∼ Normal(µ, σ2) ∼ Normal(δσ, σ2).
I δ ∼ Normal(0, σ2δ )
I Default: σ2δ = 1.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
What is the Right Value for Variance σ20
I If σ2 is small, then σ20 should be small (perception)
I If σ2 is large, then σ20 should be large (clinical applications)
I Idea: Place prior on effect size instead: δ = µ/σ.
yi ∼ Normal(µ, σ2) ∼ Normal(δσ, σ2).
I δ ∼ Normal(0, σ2δ )
I Default: σ2δ = 1.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
What is the Right Value for Variance σ20
I If σ2 is small, then σ20 should be small (perception)
I If σ2 is large, then σ20 should be large (clinical applications)
I Idea: Place prior on effect size instead: δ = µ/σ.
yi ∼ Normal(µ, σ2) ∼ Normal(δσ, σ2).
I δ ∼ Normal(0, σ2δ )
I Default: σ2δ = 1.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
−4 −2 0 2 4
0.0
0.1
0.2
0.3
0.4
0.5
Effect Size δδ
Den
sity
Null
Alternative
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Jeffreys-Zellner-Siow Prior
I Place prior distribution on σ2δ
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
0 2 4 6 8
0.0
0.1
0.2
0.3
0.4
Prior Variance σσδδ2
Den
sity
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
−5 0 5
0.0
0.1
0.2
0.3
0.4
Effect Size δδ
Den
sity
CauchyNormal
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
I We recommend the JZS as a noninformative default.
I Normal prior is defensible too
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
How to Calculate a Bayes Factor
pcl.missouri.edu/bayesfactor
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Scaling Alternative
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Scaling Alternative
−5 0 5
0.0
0.2
0.4
0.6
0.8
Effect Size δδ
Den
sity
Scale .4Scale 1.0
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Scaling Alternative
I If small effect sizes are of interest, use small scales.
I If moderate or large effects sizes are of interest, use defaultvalue of 1.0.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Sample Size
Crit
ical
t−va
lue
5 10 20 50 100 500 2000 5000
23
45
6JZS BFUnit BFBICp−value
p=.05
B=.1
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Questionable Effects
I Grider & Malberg (2008) claim emotional words are betterremembered than neutral ones.
I .76 vs. .80, t(79) = 2.24I B01 = 1.02I No evidence for either hypothesis.
I Plant & Peruche (2005) claim sensitivity training reducedshooter bias.
I F (1, 47) = 5.70I t(47) = 2.39I B01 = .66 or 1.6 : 1 in favor of alternativeI Miniscule evidence for the claim
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Questionable Effects
I Grider & Malberg (2008) claim emotional words are betterremembered than neutral ones.
I .76 vs. .80, t(79) = 2.24I B01 = 1.02I No evidence for either hypothesis.
I Plant & Peruche (2005) claim sensitivity training reducedshooter bias.
I F (1, 47) = 5.70I t(47) = 2.39I B01 = .66 or 1.6 : 1 in favor of alternativeI Miniscule evidence for the claim
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
What about really small effects (δ = .02)?
Bay
es F
acto
r
5 10 20 50 100
200
500
1000
2000
5000
1000
0
2000
0
5000
0
1e+
05
0.01
0.1
1
10
Sample Size
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Bayes Factor
I Bayes factor gives researchers a principled way of acceptingand rejecting the null
I Bayes factor requires specification of alternatives.
I The JZS prior is a good noninformative choice
I Extensions to factorial designs are coming (NIH willing)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Bayes Factor
I Bayes factor gives researchers a principled way of acceptingand rejecting the null
I Bayes factor requires specification of alternatives.
I The JZS prior is a good noninformative choice
I Extensions to factorial designs are coming (NIH willing)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Bayes Factor
I Bayes factor gives researchers a principled way of acceptingand rejecting the null
I Bayes factor requires specification of alternatives.
I The JZS prior is a good noninformative choice
I Extensions to factorial designs are coming (NIH willing)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Bayes Factor
I Bayes factor gives researchers a principled way of acceptingand rejecting the null
I Bayes factor requires specification of alternatives.
I The JZS prior is a good noninformative choice
I Extensions to factorial designs are coming (NIH willing)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Bayes Factor
I Bayes factor gives researchers a principled way of acceptingand rejecting the null
I Bayes factor requires specification of alternatives.
I The JZS prior is a good noninformative choice
I Extensions to factorial designs are coming (NIH willing)
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Hypothesis Testing
I Do hypothesis testing only if (1) it is necessary and (2) youare willing to accept the null. Alternative: Explore data forstructure.
I Hypothesis testing is necessarily subjective, but not too analarming degree.
I We have the communal infrastructure to evaluate subjectiveelements in science.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Hypothesis Testing
I Do hypothesis testing only if (1) it is necessary and (2) youare willing to accept the null. Alternative: Explore data forstructure.
I Hypothesis testing is necessarily subjective, but not too analarming degree.
I We have the communal infrastructure to evaluate subjectiveelements in science.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Hypothesis Testing
I Do hypothesis testing only if (1) it is necessary and (2) youare willing to accept the null. Alternative: Explore data forstructure.
I Hypothesis testing is necessarily subjective, but not too analarming degree.
I We have the communal infrastructure to evaluate subjectiveelements in science.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests
Conclusions: Hypothesis Testing
I Do hypothesis testing only if (1) it is necessary and (2) youare willing to accept the null. Alternative: Explore data forstructure.
I Hypothesis testing is necessarily subjective, but not too analarming degree.
I We have the communal infrastructure to evaluate subjectiveelements in science.
Jeffrey N. Rouder University of Missouri
Bayes factors as a replacement for t-tests