reading efron's 1979 paper on bootstrap
DESCRIPTION
seminar talk of Marco Brandi, Nov. 26, 2012TRANSCRIPT
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Bootstrap Methods:Another Look at the Jackknife
Marco Brandi
TSI-EuroBayes StudentUniversity Paris Dauphine
26 November 2012 / Reading Seminar on Classics
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
"To pull oneself up by one is bootstrap"
Rudolph Erich Raspe
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
OUTLINE
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
PRESENTING THE PROBLEM
X = (X1, . . . ,Xn)
Xi ∼ F with F completely unspecified
GOAL
⇓
Given R(X,F ) estimate R on the basis of x = (x1, . . . , xn)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
PRESENTING THE PROBLEM
X = (X1, . . . ,Xn)
Xi ∼ F with F completely unspecified
GOAL
⇓
Given R(X,F ) estimate R on the basis of x = (x1, . . . , xn)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
INTRODUCTION JACKKNIFE METHOD
θ(F ) parameter of interest and t(X) its estimatorR(X,F ) = t(X)− θ(F )
R(X,F ) = t(X)− ˆBias(t)−θ(F )
ˆ(Var(t))1/2
ˆBias(t) and ˆVar(t) are obtained recomputing t(·) n times , eachtime removing one component of X
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP METHOD
BOOTSTRAP METHODat x1, x2, . . . , xn put mass 1/n
F̂ is the sample probability distributionX ∗i = x∗i X ∗i ∼ F̂ i = 1, . . . ,nX∗ boostrap sampleR∗ = R(X∗, F̂ )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP METHOD
BOOTSTRAP METHODat x1, x2, . . . , xn put mass 1/nF̂ is the sample probability distribution
X ∗i = x∗i X ∗i ∼ F̂ i = 1, . . . ,nX∗ boostrap sampleR∗ = R(X∗, F̂ )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP METHOD
BOOTSTRAP METHODat x1, x2, . . . , xn put mass 1/nF̂ is the sample probability distributionX ∗i = x∗i X ∗i ∼ F̂ i = 1, . . . ,n
X∗ boostrap sampleR∗ = R(X∗, F̂ )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP METHOD
BOOTSTRAP METHODat x1, x2, . . . , xn put mass 1/nF̂ is the sample probability distributionX ∗i = x∗i X ∗i ∼ F̂ i = 1, . . . ,nX∗ boostrap sample
R∗ = R(X∗, F̂ )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP METHOD
BOOTSTRAP METHODat x1, x2, . . . , xn put mass 1/nF̂ is the sample probability distributionX ∗i = x∗i X ∗i ∼ F̂ i = 1, . . . ,nX∗ boostrap sampleR∗ = R(X∗, F̂ )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
SIMPLE EXAMPLE
Dichotomous Example
θ(F ) = Pr{X = 1} R(X,F ) = X̄ − θ(F )
{X ∗i = 1 x̄ = θ(F̂ )
X ∗i = 0 1− x̄
⇓
R∗ = R(X∗, F̂ ) = X̄ ∗ − x̄
E∗(X̄ ∗ − x̄) = 0 Var∗(X̄ ∗ − x̄) = x̄(1− x̄)/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
SIMPLE EXAMPLE
Dichotomous Example
θ(F ) = Pr{X = 1} R(X,F ) = X̄ − θ(F )
{X ∗i = 1 x̄ = θ(F̂ )
X ∗i = 0 1− x̄
⇓
R∗ = R(X∗, F̂ ) = X̄ ∗ − x̄
E∗(X̄ ∗ − x̄) = 0 Var∗(X̄ ∗ − x̄) = x̄(1− x̄)/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
SIMPLE EXAMPLE
Dichotomous Example
θ(F ) = Pr{X = 1} R(X,F ) = X̄ − θ(F )
{X ∗i = 1 x̄ = θ(F̂ )
X ∗i = 0 1− x̄
⇓
R∗ = R(X∗, F̂ ) = X̄ ∗ − x̄
E∗(X̄ ∗ − x̄) = 0 Var∗(X̄ ∗ − x̄) = x̄(1− x̄)/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
PROBLEM
The complexity on the bootstrap procedure is to calculatethe bootstrap distribution
⇓
3 methods of calculation are possible
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
PROBLEM
The complexity on the bootstrap procedure is to calculatethe bootstrap distribution
⇓
3 methods of calculation are possible
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
Method 1
Direct theoretical calculation
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 1ST STEP
Initializing the procedure
θ(F ) indicate the median of F
t(X) = X(m)
X(1) ≤ X(2) ≤ · · · ≤ X(n) n = 2m − 1R(X,F ) = t(X)− θ(F )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 1ST STEP
Initializing the procedure
θ(F ) indicate the median of Ft(X) = X(m)
X(1) ≤ X(2) ≤ · · · ≤ X(n) n = 2m − 1R(X,F ) = t(X)− θ(F )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 1ST STEP
Initializing the procedure
θ(F ) indicate the median of Ft(X) = X(m)
X(1) ≤ X(2) ≤ · · · ≤ X(n) n = 2m − 1
R(X,F ) = t(X)− θ(F )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 1ST STEP
Initializing the procedure
θ(F ) indicate the median of Ft(X) = X(m)
X(1) ≤ X(2) ≤ · · · ≤ X(n) n = 2m − 1R(X,F ) = t(X)− θ(F )
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 2ST STEP
Formalazing the procedureX∗ = x∗
N∗i = #{X ∗i = xi} N∗ = (N∗1 ,N∗1 , . . . .N
∗n)
R∗ = R(X∗, F̂ ) = X ∗(m) − x(m)
Pr∗{R∗ = x(l) − x(m)} =Pr{Bin(n,l − 1
n) ≤ m − 1}−
−Pr{Bin(n,ln
) ≤ m − 1}(1)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 2ST STEP
Formalazing the procedureX∗ = x∗
N∗i = #{X ∗i = xi} N∗ = (N∗1 ,N∗1 , . . . .N
∗n)
R∗ = R(X∗, F̂ ) = X ∗(m) − x(m)
Pr∗{R∗ = x(l) − x(m)} =Pr{Bin(n,l − 1
n) ≤ m − 1}−
−Pr{Bin(n,ln
) ≤ m − 1}(1)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 2ST STEP
Formalazing the procedureX∗ = x∗
N∗i = #{X ∗i = xi} N∗ = (N∗1 ,N∗1 , . . . .N
∗n)
R∗ = R(X∗, F̂ ) = X ∗(m) − x(m)
Pr∗{R∗ = x(l) − x(m)} =Pr{Bin(n,l − 1
n) ≤ m − 1}−
−Pr{Bin(n,ln
) ≤ m − 1}(1)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
ESTIMATING THE MEDIAN 2ST STEP
Formalazing the procedureX∗ = x∗
N∗i = #{X ∗i = xi} N∗ = (N∗1 ,N∗1 , . . . .N
∗n)
R∗ = R(X∗, F̂ ) = X ∗(m) − x(m)
Pr∗{R∗ = x(l) − x(m)} =Pr{Bin(n,l − 1
n) ≤ m − 1}−
−Pr{Bin(n,ln
) ≤ m − 1}(1)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
RESULTS(1)
for n = 15 and m = 8l 2 or 14 3 or 13 4 or 12 5 or 11 6 or 10 7 or 9 8
(1) .0003 .0040 .0212 .0627 .1249 .1832 .2073
Use E∗(R∗)2 =∑15
l=1[x(l) − x(8)]2Pr∗{
R∗ = x(l) − x(8)}
as an estimate of EF R2 = EF [t(X)− θ(F )]2
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
RESULTS(2)
Results for bootstrap
limn→∞ nE∗(R∗)2 = 1/4f 2(θ)
Results for the standard jackknife
limn→∞ n ˆVar(R) = (1/4f 2(θ))[χ2
2]2
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
RESULTS(2)
Results for bootstrap
limn→∞ nE∗(R∗)2 = 1/4f 2(θ)
Results for the standard jackknife
limn→∞ n ˆVar(R) = (1/4f 2(θ))[χ2
2]2
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
METHOD 2 - MONTE CARLO APPROXIMATION
Repeat X∗ B times
x∗1,x∗2, . . . ,x∗B
R(x∗1, F̂ ),R(x∗2, F̂ ), . . . ,R(x∗B, F̂ )
is taken as an approximation of the boostrap distribution
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(1)
Xi ∼ Pois(2) i = 1, . . . ,15
t(X) = E [X]
B = 10000n◦of bootstrap samplesmean = 1.9341se = 0.382
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(1)
Xi ∼ Pois(2) i = 1, . . . ,15
Histogram of bootstrap mean
Bootstrap estimation of mean
De
nsity
0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.4
0.8
t(X) = E [X]
B = 10000n◦of bootstrap samplesmean = 1.9341se = 0.382
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(1)
Xi ∼ Pois(2) i = 1, . . . ,15
Histogram of bootstrap mean
Bootstrap estimation of mean
De
nsity
0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.4
0.8
t(X) = E [X]
B = 10000n◦of bootstrap samples
mean = 1.9341se = 0.382
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(1)
Xi ∼ Pois(2) i = 1, . . . ,15
Histogram of bootstrap mean
Bootstrap estimation of mean
De
nsity
0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.4
0.8
t(X) = E [X]
B = 10000n◦of bootstrap samplesmean = 1.9341
se = 0.382
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(1)
Xi ∼ Pois(2) i = 1, . . . ,15
Histogram of bootstrap mean
Bootstrap estimation of mean
De
nsity
0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.0
0.4
0.8
t(X) = E [X]
B = 10000n◦of bootstrap samplesmean = 1.9341se = 0.382
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(2)
t(X) = V [X]
B = 10000n◦of bootstrap samplesmean = 2.191se = 0.649
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(2)
Histogram of bootstrap variance
Bootstrap estimation of variance
De
nsity
0 1 2 3 4 5
0.0
0.2
0.4
t(X) = V [X]
B = 10000n◦of bootstrap samplesmean = 2.191se = 0.649
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(2)
Histogram of bootstrap variance
Bootstrap estimation of variance
De
nsity
0 1 2 3 4 5
0.0
0.2
0.4
t(X) = V [X]
B = 10000n◦of bootstrap samples
mean = 2.191se = 0.649
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(2)
Histogram of bootstrap variance
Bootstrap estimation of variance
De
nsity
0 1 2 3 4 5
0.0
0.2
0.4
t(X) = V [X]
B = 10000n◦of bootstrap samplesmean = 2.191
se = 0.649
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
EXAMPLE(2)
Histogram of bootstrap variance
Bootstrap estimation of variance
De
nsity
0 1 2 3 4 5
0.0
0.2
0.4
t(X) = V [X]
B = 10000n◦of bootstrap samplesmean = 2.191se = 0.649
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
R CODE
## s imu la t i on poisson dataset . seed (592)x= rpo i s (15 , lambda=2)B=10000## create the boots t rap f u n c t i o nboots t rap <− f u n c t i o n ( data , nboot , theta , . . . ){
z <− l i s t ( )datab <−
mat r i x ( sample ( data , s i ze= leng th ( data )∗nboot , rep lace=TRUE) , nrow=nboot )estb <− apply ( datab ,1 , theta , . . . )es t <− t he ta ( data , . . . )z$est <− estz$ d i s t n <− estbz$bias <− mean( estb)−estz$se <− sd ( estb )z
}## Est imat ing the meanX1=boots t rap ( x ,B, the ta=mean)h i s t (X1$ d is tn , main=" Histogram of boo ts t rap mean" , prob=T ,x lab=" Boots t rap es t ima t ion o f mean" )mean(X1$ d i s t n )X1$se
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
METHOD 3 - RELATIONSHIP WITH THE JACKKNIFE
P∗i = N∗i /n P∗ = (P∗1 ,P∗2 , . . . ,P
∗n)
E∗P∗ = e/n Cov∗P∗ = I/n2 − e′e/n3
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
USING TAYLOR EXPANSION
R(P∗) = R(X∗, F̂ ) evaluate in P∗ = e/n
R(P∗) = R(e/n) + (P∗ − e/n)U +12
(P∗ − e/n)V(P∗ − e/n)′
U =
...
∂R(P∗)∂P∗
i...
P∗=e/n
V =
...
......
... ∂2R(P∗)∂P∗
i ∂P∗j
......
......
P∗=e/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
USING TAYLOR EXPANSION
R(P∗) = R(X∗, F̂ ) evaluate in P∗ = e/n
R(P∗) = R(e/n) + (P∗ − e/n)U +12
(P∗ − e/n)V(P∗ − e/n)′
U =
...
∂R(P∗)∂P∗
i...
P∗=e/n
V =
...
......
... ∂2R(P∗)∂P∗
i ∂P∗j
......
......
P∗=e/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
USING TAYLOR EXPANSION
R(P∗) = R(X∗, F̂ ) evaluate in P∗ = e/n
R(P∗) = R(e/n) + (P∗ − e/n)U +12
(P∗ − e/n)V(P∗ − e/n)′
U =
...
∂R(P∗)∂P∗
i...
P∗=e/n
V =
...
......
... ∂2R(P∗)∂P∗
i ∂P∗j
......
......
P∗=e/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
DERIVATION OF BOOTSTRAP EXPECTATION ANDVARIANCE
R(P∗) = R(
P∗∑ni=1 P∗i
)eU = 0 eV = −nU′ eVe′ = 0
E∗R(P∗) = R(e/n) +12
trV[I/n2 − e′e/n3
]= R(e/n) +
12n
V̄
Var∗R(P∗) = U′[I/n2 − e′e/n3
]U =
n∑i=1
U2i /n
2
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
DERIVATION OF BOOTSTRAP EXPECTATION ANDVARIANCE
R(P∗) = R(
P∗∑ni=1 P∗i
)eU = 0 eV = −nU′ eVe′ = 0
E∗R(P∗) = R(e/n) +12
trV[I/n2 − e′e/n3
]= R(e/n) +
12n
V̄
Var∗R(P∗) = U′[I/n2 − e′e/n3
]U =
n∑i=1
U2i /n
2
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
METHOD 1METHOD 2METHOD 3
RESULTS
BiasFθ(F̂ ) ≈ 12n V̄
VarFθ(F̂ ) ≈∑n
i=1 U2i /n
2
The results agree with those given by Jaeckel’s infinitesimaljackknife
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
REGRESSION MODELS
Xi = gi(β) + εi εi ∼ F i = 1, . . . ,n
Having observed X = x we compute the estimate of β
β̂ = minβn∑
i=1
[xi − gi
(β̂)]2
F̂ : mass1n
at ε̂i = xi − gi
(β̂)
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BOOTSTRAP SAMPLE
X ∗i = gi
(β̂)
+ ε∗i ε∗i ∼ F̂
β̂∗ : minβn∑
i=1
[x∗i − gi
(β̂)]2
β̂∗1, β̂∗2, β̂∗3, . . . , β̂∗B
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
LINEAR MODEL
gi(β) = ciβ C′C = G
β̂ = G−1C′X has mean β and covariance matrix σ2F G−1
β̂∗ = G−1C′X∗ has boostrap mean and variance
E∗β̂∗ = β̂ Cov∗β̂∗ = σ̂2G−1
where σ̂2 =∑n
i=1
[xi − g
(β̂)]2
/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
LINEAR MODEL
gi(β) = ciβ C′C = G
β̂ = G−1C′X has mean β and covariance matrix σ2F G−1
β̂∗ = G−1C′X∗ has boostrap mean and variance
E∗β̂∗ = β̂ Cov∗β̂∗ = σ̂2G−1
where σ̂2 =∑n
i=1
[xi − g
(β̂)]2
/n
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
JACKKNIFE IN LINEAR REGRESSION
Applying the infinitesimal jackknife in a linear regression model,Hinkley derive the approximation of
Cov β̂ ≈ G−1
[n∑
i=1
c′i ci ε̂2i
]G−1
Jackknife methods ignore that the errors εi are assumed tohave the same distribution for every value of i
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
DEFINITION OF BAYESIAN BOOTSTRAP (D. Rubin1981)
Bayesian BootstrapIn bootstrap we consider sample cdf is population cdf
Each BB replications generates a posterior probability foreach xi
The posterior probability of each xi is centered at 1n but has
variability
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
DEFINITION OF BAYESIAN BOOTSTRAP (D. Rubin1981)
Bayesian BootstrapIn bootstrap we consider sample cdf is population cdfEach BB replications generates a posterior probability foreach xi
The posterior probability of each xi is centered at 1n but has
variability
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
DEFINITION OF BAYESIAN BOOTSTRAP (D. Rubin1981)
Bayesian BootstrapIn bootstrap we consider sample cdf is population cdfEach BB replications generates a posterior probability foreach xi
The posterior probability of each xi is centered at 1n but has
variability
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BB REPLICATION
BB replication
(n − 1) Unif (0,1) u(0) = 0 e u(n) = 1
gl = u(l) − u(l−1)
Attach the vector (g1, . . . ,gn) to the data X
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BB REPLICATION
BB replication
(n − 1) Unif (0,1) u(0) = 0 e u(n) = 1gl = u(l) − u(l−1)
Attach the vector (g1, . . . ,gn) to the data X
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BB REPLICATION
BB replication
(n − 1) Unif (0,1) u(0) = 0 e u(n) = 1gl = u(l) − u(l−1)
Attach the vector (g1, . . . ,gn) to the data X
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
CONCEPTUAL DIFFERENCE
Bayesian BootstrapSimulates the posterior distribution of the parameter
Classical BootstrapSimulates the estimated sampling distribution of a statistic
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
CONCEPTUAL DIFFERENCE
Bayesian BootstrapSimulates the posterior distribution of the parameter
Classical BootstrapSimulates the estimated sampling distribution of a statistic
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BB EXAMPLE
Dichotomous Example
The parameter is θ = Pr{Xi = 1} and let n1 number of Xi = 1
Call P1 the sum of the n1 probabilities assigned to the xi = 1
(g1, . . . ,gn) ∼ Dirichlet(1, . . . ,1)⇒ P1 ∼ Beta(n1,n − n1)
Note: Beta(n1,n − n1) is the posterior distribution when theprior is P(θ) ∝ [θ(1− θ)]−1
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
INFERENCES PROBLEMS
Is it possible that all the values of X have been observed?Is it reasonable to assume a priori independentparameters, constrained only to sum to 1, for these values?
Using the gap to simulate the posterior distributions ofparameters may no longer work
so..BB and bootstrap cannot avoid the sensitivity of inference to
model assumptions
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
INFERENCES PROBLEMS
Is it possible that all the values of X have been observed?
Is it reasonable to assume a priori independentparameters, constrained only to sum to 1, for these values?
Using the gap to simulate the posterior distributions ofparameters may no longer work
so..BB and bootstrap cannot avoid the sensitivity of inference to
model assumptions
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
INFERENCES PROBLEMS
Is it possible that all the values of X have been observed?Is it reasonable to assume a priori independentparameters, constrained only to sum to 1, for these values?
Using the gap to simulate the posterior distributions ofparameters may no longer work
so..BB and bootstrap cannot avoid the sensitivity of inference to
model assumptions
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
INFERENCES PROBLEMS
Is it possible that all the values of X have been observed?Is it reasonable to assume a priori independentparameters, constrained only to sum to 1, for these values?
Using the gap to simulate the posterior distributions ofparameters may no longer work
so..BB and bootstrap cannot avoid the sensitivity of inference to
model assumptions
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
CONCLUSION
Knowledge of the context of a data set may make theincorporation of reasonable model constraints obvious and
bootstrap may be useful in particular contexts
In general"There are no general data analytic panaceas thatallow us to pull ourselves up by our bootstraps"
Donald Rubin
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
CONCLUSION
Knowledge of the context of a data set may make theincorporation of reasonable model constraints obvious and
bootstrap may be useful in particular contexts
In general"There are no general data analytic panaceas thatallow us to pull ourselves up by our bootstraps"
Donald Rubin
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
Outline
1 INTRODUCTION
2 DESCRIPTION OF METHODSMETHOD 1METHOD 2METHOD 3
3 BOOTSTRAP IN REGRESSION MODELS
4 BAYESIAN BOOTSTRAP
5 DISCUSSION
6 BAG OF LITTLE BOOTSTRAP
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BLB (M. Jordan 2012)
When n gets large computational cost is largeExpected numbers of distinct points in a resample is ∼ 0.632n
BLB ProcedureDivide the dataset in s subset of dimension b, with b < n
From each subset we draw r samples with replacement ofdimension nCompute for each subset the estimator quality assessment(e.g the bias) indicated with ξ
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BLB (M. Jordan 2012)
When n gets large computational cost is largeExpected numbers of distinct points in a resample is ∼ 0.632n
BLB ProcedureDivide the dataset in s subset of dimension b, with b < nFrom each subset we draw r samples with replacement ofdimension n
Compute for each subset the estimator quality assessment(e.g the bias) indicated with ξ
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BLB (M. Jordan 2012)
When n gets large computational cost is largeExpected numbers of distinct points in a resample is ∼ 0.632n
BLB ProcedureDivide the dataset in s subset of dimension b, with b < nFrom each subset we draw r samples with replacement ofdimension nCompute for each subset the estimator quality assessment(e.g the bias) indicated with ξ
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
BLB IMAGE
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
FINALLY...
if we choose b = n0.6 ad we have a dataset of 1TB, thesubsamples contains at most 3981 distinct points and have sizeat most 4GB
Like the bootstrapShare bootstrap’s consistencyAutomatic : without knowledge of the internals θ
Beyond the bootstrapCan explicity control b
Generally faster than the bootstrap and requires less totalcomputation
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
FINALLY...
if we choose b = n0.6 ad we have a dataset of 1TB, thesubsamples contains at most 3981 distinct points and have sizeat most 4GB
Like the bootstrapShare bootstrap’s consistencyAutomatic : without knowledge of the internals θ
Beyond the bootstrapCan explicity control bGenerally faster than the bootstrap and requires less totalcomputation
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
References I
B. Efron.Bootstrap Methods: Another Look at the Jackknife.The Annals of Statistics, Vol. 7, No. 1, (Jan. 1979), pp. 1-26.
D.B. Rubin.The Bayesian Bootstrap.The Annals of Statistics, Vol. 9, No.1, pp. 130-134.
M. Jordan.The Big Data Bootstrap.Proceedings of the 29th International Conference onMachine Learning (ICML).
Marco Brandi Bootstrap Methods: Another Look at the Jackknife
INTRODUCTIONDESCRIPTION OF METHODS
BOOTSTRAP IN REGRESSION MODELSBAYESIAN BOOTSTRAP
DISCUSSIONBAG OF LITTLE BOOTSTRAP
THANK YOU
FOR
YOUR ATTENTION
Marco Brandi Bootstrap Methods: Another Look at the Jackknife