7
The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of Moments Generalized Method of Moments Topic: parametric estimation of the invariants distribution based on the Generalized Method of Moments We introduce the Method of Moments with Flexible Probabilities estimate We extend to Flexible Probabilities the Generalized Method of Moments with exact specification ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

Embed Size (px)

Citation preview

Page 1: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of Moments

Generalized Method of Moments

• Topic: parametric estimation of the invariants distribution based onthe Generalized Method of Moments

• We introduce the Method of Moments with FlexibleProbabilities estimate

• We extend to Flexible Probabilities the Generalized Method ofMoments with exact specification

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 2: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsMethod of Moments

Canonical Method of Moments

A parametric assumption on the distribution of the invariants is

εt ∼ fε ∈ {fθ}θ∈Θ (2a.95)

• θ ≡ (θ1, . . . , θl̄)′, l̄ × 1 vector

• Θ ⊆ Rl̄, discrete or continuous

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 3: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsMethod of Moments

Canonical Method of Moments

The first l̄ cross moments of the invariants can be expressed by l̄ equations

fε ≡ fθε , E {εi,tεj,t · · · } ≡ hi,j,...(θε) (2a.96)

The Method of Moments (MM) estimate θ̂MM

ε of θε is the solution of

f̂MMε ≡ f

θ̂MMε

,1

∑t̄t=1εi,tεj,t · · · ≡ hi,j,...(θ̂

MM

ε ) (2a.101)

where {ε1, . . . , εt̄} is the time series of realized invariants.

See also Example 2a.24

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 4: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsMethod of Moments

Method of Moments with Flexible Probabilities

Consider Flexible Probabilities {pt}t̄t=1 rather than the equal probabilityweights pt ≡ 1/t̄.

The Method of Moments with Flexible Probabilities (MMFP)estimate θ̂

MMFP

ε of θε is

f̂MMFPε ≡ f

θ̂MMFPε

,∑t̄

t=1ptεi,tεj,t · · · ≡ hi,j,...(θ̂MMFP

ε ) (2a.102)

See also Example 2a.25

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 5: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsMethod of Moments

Example 2a.25. Fit of options strategy toreflected-shifted-lognormal distribution using MMFP

• Invariant: daily P&L εt ≡ Πt−1→t modeled as a reflected shiftedlognormal (2a.91)

• FP: time exponential decay probabilities, half life τHL = 500

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 6: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsGeneralized Method of Moments - exact specification

Exact specification

Conditions on a set of l̄ arbitrary moment or orthogonality con-ditions

fε ≡ fθε , E {g (εt,θε)} ≡ 0 (2a.104)

• g (x,θ) = (g1 (x,θ) , . . . , gl̄ (x,θ))′ multivariate function

• θε ∈ Θ ⊆ Rl̄ vector of true unknown parameters

This problem is exactly specified, i.e. l̄ equations in l̄ unknowns.

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update

Page 7: ARPM _ The “Checklist” - 2a. Estimation Flexible Probabilities - Generalized Method of Moments

The “Checklist” > 2a. Estimation: Flexible Probabilities > Generalized Method of MomentsGeneralized Method of Moments - exact specification

Exact specification

The exact-specified Generalized Method of Moments with FlexibleProbabilities (GMMFP) estimate θ̂

GMMFP

ε of θε is the solution of

f̂GMMFPε ≡ f

θ̂GMMFPε

,∑t̄

t=1ptg(εt, θ̂GMMFP

ε ) ≡ 0 (2a.106)

Consistency: the parametric assumption (2a.95) holding true andprovided θε is the unique solution of (2a.108) (respectively, (2a.104) forexact-specified GMMFP estimate)

limens(p)→∞ θ̂GMMFP

ε = θε ⇔ limens(p)→∞ f̂GMMFPε = fε (2a.116)

where ens(p) is the Effective Number of Scenarios (2a.21).

See also Example 2a.26 and Example 2a.27

ARPM - Advanced Risk and Portfolio Management - arpm.co This update: Feb-01-2017 - Last update