21MLE(1)

# 21MLE(1) - c 2011Dept.Statistics(IowaStateUniversity...

This preview shows pages 1–3. Sign up to view the full content.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: MAXIMUMLIKELIHOODandREMLESTIMATION INTHEGENERALLINEARMODEL c 2011Dept.Statistics(IowaStateUniversity) Stat511section21 1/23 Suppose f ( w | θ ) istheprobabilitydensityfunction ( pdf ) or probabilitymassfunction ( pmf ) ofarandomvector w ,where θ isa k × 1 vectorofparameters. Givenavalueoftheparametervector θ , f ( w | θ ) isareal-valued functionof w . Thelikelihoodfunction L ( θ | w )= f ( w | θ ) isareal-valuedfunctionof θ foragivenvalueof w . L ( θ | w ) isnotapdf. θ L ( θ | w ) d θ = 1 . Foranypotentialobservedvalue w ,define ˆ θ ( w ) tobeaparameter valueatwhich L ( θ | w ) attainsitsmaximumvalue. ˆ θ ( w ) isa maximumlikelihoodestimator (MLE)of θ . InvariancepropertyofMLES:TheMLEofafunctionof θ ,say g ( θ ) ,isthefunctionevaluatedattheMLEof θ : g ( θ ) = g ( ˆ θ ) c 2011Dept.Statistics(IowaStateUniversity) Stat511section21 2/23 Oftenmuchmoreconvenienttoworkwith l ( θ | w )= log L ( w | θ ) . If l ( θ | w ) isdifferentiable,candidatesfortheMLEof θ canbefound byequatingthescorefunction ∂ l ( θ | w ) ∂ θ ≡ ⎡ ⎢⎢ ⎣ ∂ l ( θ | w ) ∂ θ .. . ∂ l ( θ | w ) ∂ θ k ⎤ ⎥⎥ ⎦ to andsolvingfor θ Thescoreequationsare ∂ l ( θ | w ) ∂ θ = ⇒ ∂ l ( θ | w ) ∂θ j = ∀ j = 1 ,..., k OnestrategyforobtaininganMLEistofindsolution(s)ofthe scoreequationsandverifythatatleastonesuchsolution maximizes l ( θ | w ) . Ifthesolution(s)tothescoreequationslieoutsidetheappropriate parameterspace,theyarenotMLE’s c 2011Dept.Statistics(IowaStateUniversity) Stat511section21 3/23 Example:NormalTheoryGauss-MarkovLinearModel y n × 1 = X n × p β p × 1 + n × 1 ∼ N ( ,σ 2 I ) θ ( p + 1 ) × 1 = β σ 2 f ( y | θ )= exp { − 1 2 ( y − x β ) ( σ 2 I ) − 1 ( y − x β ) } ( 2 π ) n / 2 | σ 2 I | 1 / 2 = 1 ( 2 πσ 2 ) n / 2 exp − 1 2 σ 2 ( y − x β ) ( y − x β ) l ( θ | y )= − n 2 log2 πσ 2 − 1 2 σ 2 ( y − x β ) ( y − x β ) Thescorefunctionis ∂ l ( θ | y ) ∂ θ = ∂ l ( θ | y ) ∂ β ∂ l ( θ | y ) ∂σ 2 = 1 σ 2 ( x y − x x β ) ( y − x β ) ( y − x β ) 2 σ 4 − n 2 σ 2 Thescoreequationsare ∂ l ( θ | y ) ∂ θ = ⇔ x x β = x y σ 2 = ( y − x β ) ( y − x β ) n c 2011Dept.Statistics(IowaStateUniversity) Stat511section21 4/23 Anysolutionliesintheappropriateparameterspace: β ∈ IR p , σ 2 ∈ IR + . Asolutiontothescoreequationsis ˆ β ( y − x ˆ β ) ( y − x ˆ β ) n , where ˆ β isasolutiontothenormalequations Needtoshowthisisamaximumofthelikelihoodfunction Wealreadyknowthatanysolutiontothenormalequations, β = ˆ β ,minimizes ( y − X β ) ( y − X β ) for β ∈ IR p . Thus, ∀ σ 2 > , l ˆ β σ 2 | y ≥ l β σ 2 | y ∀...
View Full Document

{[ snackBarMessage ]}

### Page1 / 6

21MLE(1) - c 2011Dept.Statistics(IowaStateUniversity...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online