regression analysis -...

48
Regression Analysis ~Basic concept ~ 迴歸分析基本觀念

Upload: duongtram

Post on 01-Sep-2018

296 views

Category:

Documents


0 download

TRANSCRIPT

  • Regression Analysis

    Basic concept

  • Regression Analysis

    1805(Legendre)1809(Gauss)

    Francis Galton regression to the mean

  • Types of Relationships

    Y

    X

    Y

    X

    Y

    Y

    X

    X

    Linear relationships Curvilinear relationships

    DCOVA

    Copyright 2014 Statistics for Managers Using Microsoft Excel 7th Edition, Global Edition

  • Types of Relationships(continued)

    Y

    X

    Y

    X

    Strong relationships

    Y

    Y

    X

    X

    Weak relationships

    DCOVA

    Copyright 2014 Statistics for Managers Using Microsoft Excel 7th Edition, Global Edition

    Y

    X

    Y

    X

    No relationship

  • independent variableX

    dependent variable Y

    explanatory variable

    response variable

  • Y

    (X)

    simple regression analysisy=fx

    multiple regression analysisy=fx1, x2, x3, ., xn

    multivariate regression analysisy1, y2, .., ym=fx1, x2, ., xn

    linear regression

    nonlinear regression

  • simple regression analysis

  • xi

    yi

    xiyi

    = 0 + 1

    fy xi

    = 0 + 1

  • xix1

    = = 0 + 1

    fy xi

    yxiyx1

    xi

    x1

    50 kg

    50kg170 cm

  • 1.y

    2.homocedasticityy

    Var(y)=12 = 2

    2 = = 2 =

    3.

    = 0 + 1 01

    4.XY

    5.0

    N0, 2 Cov , j 0

  • ()

    xix1

    = 0 + 1

    ( xi , yi )

    (xi , i )

    ( xi , yi ) 00 11

  • Y

    XXi

    i

    ( xi , yi )

    (xi , i )

    =

    = 0 + 1

    ()i Xi

  • i

    =

    min2 = min

    2 = min 0 12 = min (SSE)

  • min 0 12= min (SSE) 0 1

    0

    SSE

    0= 0

    SSE

    1= 0

    2 0 1(1) = 0

    2 0 1() = 0

    = 0 + 1 (1)

    = 0 + 12

    1 =( )( )

    ( )2

    0 =

    2

    2 ()

    2

    =

    2 2

    =

    2

    Sxy XYSx

    2 X

    1

    (1) 0 = 1 x

    y 0

  • 1. 0 1 0 1 E 0 0

    E 1 1

    2. 0 1 2

    Var 0

    2

    2

    2 Var 12

    2

    3. 0 1

    0~ 0,

    2

    2

    2 1~ 1,2

    2

    4. 0 1

  • 1

    ()

    (F)

  • (Coefficient of Determination)

    Y

    XXi

    i

    ( xi , yi )

    (xi , i )

    ei

    y = y +

    =

    = 0 + 1

    Coefficient of Determinationr

  • y2 = y +

    2

    y2 = y

    2 + 2

    SST

    SSR

    SSE

    R y2 = 0 + 1

    2= 1 + 1

    2

    12

    2 = 12xi

    2 nx2

    T y2 =yi

    2 ny2 = n 1 y2 = n y

    2

    SSESST - SSR

    = 12 n 1

    1

    1

    2 = 12 n 1

    2 = 12 n

    2

  • SSESSR

    2 =

    = 1

    R2

    2 =

    2

    2 =

    0 + 1 2

    2 =

    1 + 1 2

    2 =

    12

    2

    2

    =12

    2 2

    2 2

    =12 1 1

    2

    1 1

    2

    = 1

    2

    = 122

    2 =

    2

    2

    Coefficient of Determinationr

  • Examples of Approximate R2 Values

    R2 = 1

    Y

    X

    Y

    X

    R2 = 1

    R2 = 1

    Perfect linear relationship

    between X and Y:

    100% of the variation in Y is

    explained by variation in X

    DCOVA

    Coefficient of Determinationr

  • Examples of Approximate R2 Values

    Y

    X

    Y

    X

    0 < R2 < 1

    Weaker linear relationships

    between X and Y:

    Some but not all of the

    variation in Y is explained

    by variation in X

    DCOVA

    Coefficient of Determinationr

  • Examples of Approximate R2 Values

    R2 = 0

    No linear relationship

    between X and Y:

    The value of Y does not

    depend on X. (None of the

    variation in Y is explained

    by variation in X)

    Y

    XR2 = 0

    DCOVA

    Coefficient of Determinationr

  • F

    H01 = 0 ()

    H11 0 ()

    xy

    SSTSSR()SSE()

  • F

    F

    SSR 1 MSR=SSR/1

    F*=SR

    SE SSE n-2 MSE=SSE/(n-2)

    SST n-1

    F* F, 1, n-2 H0

    MSRSSRH0

  • R2F

    MSRSSRH0R

    2

    2 =

    F*=SR

    SE=

    SSR 1

    SSE n2

    F*=SR

    SE=

    SSR 1

    SSE n2=

    n2SSE

    SSTSSE

    SST

    =(n2)R2

    SSTSSR

    SST

    =(n2)R2

    1R2

    =R2

    1

    (1 R2)

    n 2

    R2H0R2FH0 R

    2F

  • 2

  • Y

    XXi

    i

    ( xi , yi )

    (xi , i )

    1 0

    Slope = 1 =

    (0, 0)

    Intercept

    b

    a

    = 0 + 1

  • 1

    H01 = 0 ()

    H11 0 ()

    1~ 1,2

    2 2

    t

    t =1 1S1

    =1

    MSE xi x

    2

    t*t/2, n-2 H0XY

  • 1

    1~ 1,2

    2 2

    t1-

    t =1 1S1

    =1

    MSE xi x

    2

    t1-

    1 t2,n2

    MSE

    xi x2 1 1 + t

    2,n2

    MSE

    xi x2

  • 1

    H01 0 ()

    H11 > 0 ()

    1~ 1,2

    2 2

    t

    t =1 1S1

    =1

    MSE xi x

    2

    t*t, n-2 H0XY

  • 1

    H01 0 ()

    H11 < 0 ()

    1~ 1,2

    2 2

    t

    t =1 1S1

    =1

    MSE xi x

    2

    t*- t, n-2 H0XY

  • 0

    H00 = 0 ()

    H10 0 ()

    2

    t

    t =0 0S0

    =0

    2

    2

    t*t/2, n-2 H0

    0~ 0,

    2

    22

  • 0

    2

    t1-

    t1-

    0 t2,n2

    2

    2 0 0 + t

    2,n2

    2

    2

    0~ 0,

    2

    22

    t =0 0S0

    =0

    2

    2

  • Y

    X

    = 0 + 1

    50kg

    50 kg

    50kg170 cm

    50kg

  • = 0 + 1

    fy xi

    xi

    (xi , i )

  • xy

    Eyxixi

    t2,df

    = t2, 2

    1

    +

    2

    2

    = 0 + 1 xi

    ~ ,1

    +

    2

    2

    2

    (MSE2)

  • xy

    yixi

    = t2, 2

    1 +1

    +

    2

    2

    = 0 + 1 xi

    (y)~ 0 , 1 +1

    +

    2

    2

    2

    (y)~ E(y), Var()

  • 1.

    = t2, 2

    1

    +

    2

    2

    2.MSE

    3. xi

    4. n

    5. xi

    http://faculty.cas.usf.edu/mbrannick/regression/Prediction.html

  • (Correlation Analysis)

    3

  • (Correlation Analysis)

    =

    =

    2

    2

    = ,

    =

    1 1

  • (correlation coefficient)

    1.xyxy = 0

    2. xy = 0 xy

    xy

    3.xy = 1

    4.xy = 1

  • =

    2

    2=

    1. rxy xy

    2. xy = 0 xy

    3. 2 = 2

    4. rxy = R2 1

  • 1 =

    2

    =

    1 =

    2 =

    =

    = 1 xy

    2 = 122

    2

    2 = 1

    2

    = 2

    = 2 1

  • ()

    0 = 0

    1 0

    =

    1 2

    2

    t > t2,n2 H0

    t =1

    MSE xi x

    2

  • The End