03EstimationCb(1)

# 03EstimationCb(1) - β y β Rememberourt-testquestionsabout...

This preview shows pages 1–2. 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: EstimatingEstimableFunctionsof β WenowshiftattentionfromE ( y ) totheparametervector β . Rememberourt-testquestionsabout μ 1 − μ 2 and ¯ y 1 − ¯ y 2 ? Thosearequestionsabout β orlinearcombinationsof β . We’veseensomemodelswherethereisauniquesolutionfor β sotheelementsof β canbeinterpreted.We’veseenothermodels wherethereareaninfinitenumberofsolutionsfor β . Thereareothermodelswheresomecomponentsof β have uniquesolutionsandothercomponentshaveaninfinitenumberof solutions. Howcanwetellwhenacomponentof β isunique? Howcanwetellwhenalinearcombinationof β isunique? Whatcanwesayaboutthestatisticalpropertiesofsolutionsfor β ? Copyright c 2011Dept.ofStatistics(IowaStateUniversity) Statistics511 1/19 TheResponseDependson β Onlythrough X β IntheGauss-MarkovorNormalTheoryGauss-MarkovLinear Model,thedistributionof y dependson β onlythrough X β ,i.e., y ∼ ( X β ,σ 2 I ) or y ∼ N ( X β ,σ 2 I ) If X isnotoffullcolumnrank,thereareinfinitelymanyvectorsin theset { b : Xb = X β } foranyfixedvalueof β . Thus,nomatterwhatthevalueof E ( y ) ,therewillbeinfinitelymany vectors b suchthat Xb = E ( y ) when X isnotoffullcolumnrank. Theresponsevector y canhelpuslearnabout E ( y )= X β ,but when X isnotoffullcolumnrank,thereisnohopeoflearning about β aloneunlessadditionalinformationabout β isavailable. Copyright c 2011Dept.ofStatistics(IowaStateUniversity) Statistics511 2/19 TreatmentEffectsModel Rememberthedairycowexperiment:sixexperimentalunitsrandomly assignedtotwotreatments.Onepossiblemodelfortheresponseis calledan effectsmodel . y ij = μ + τ i + ij , i = 1 , 2 ; j = 1 , 2 , 3 ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ y 11 y 12 y 13 y 21 y 22 y 23 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ = ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ μ + τ 1 μ + τ 1 μ + τ 1 μ + τ 2 μ + τ 2 μ + τ 2 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ + ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ 11 12 13 21 22 23 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ y 11 y 12 y 13 y 21 y 22 y 23 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ = ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ 110 110 110 101 101 101 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ ⎡ ⎣ μ τ 1 τ 2 ⎤ ⎦ + ⎡ ⎢⎢ ⎢⎢ ⎢⎢ ⎣ 11 12 13 21 22 23 ⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎦ Copyright c 2011Dept.ofStatistics(IowaStateUniversity) Statistics511 3/19 TreatmentEffectsModel(continued) Inthiscase,itmakesnosensetoestimate β =[ μ,τ 1 ,τ 2 ] because therearemultiple(infinitelymany,infact)choicesof β thatdefine thesamemeanfor y . Forexample, ⎡ ⎣ μ τ 1 τ 2 ⎤ ⎦ = ⎡ ⎣ 5 − 1 1 ⎤ ⎦ , ⎡ ⎣ 4 6 ⎤ ⎦ , or ⎡ ⎣ 999 − 995 − 993 ⎤ ⎦ allyieldsame X β = E ( y ) . Whenmultiplevaluesfor β definethesame E ( y ) ,wesaythat β is non-estimable . Copyright c 2011Dept.ofStatistics(IowaStateUniversity)2011Dept....
View Full Document

{[ snackBarMessage ]}

### Page1 / 5

03EstimationCb(1) - β y β Rememberourt-testquestionsabout...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online