+ All Categories
Home > Documents > Independent Dummy Variables

Independent Dummy Variables

Date post: 06-Jul-2018
Category:
Upload: m
View: 225 times
Download: 3 times
Share this document with a friend

of 32

Transcript
  • 8/17/2019 Independent Dummy Variables

    1/32

    11

    Regression when some of theRegression when some of the

    regressors are qualitativeregressors are qualitative

    Salary =f( education, experience,Salary =f( education, experience, sex, racesex, race ))

    +E+E ain!charges of "ehicle=f( #ge,ain!charges of "ehicle=f( #ge, qualityquality) +E) +E

  • 8/17/2019 Independent Dummy Variables

    2/32

    22

    $uantitative varia%les in$uantitative varia%les in

    regressionregression

    (&ummy "aria%les)(&ummy "aria%les) 'n regression analysis response varia%le is not only inuenced %y'n regression analysis response varia%le is not only inuenced %y

    the quantitative varia%les %ut also %y many varia%les of interestthe quantitative varia%les %ut also %y many varia%les of interestthat are not quantitative %ut are qualitative such as gender, race,that are not quantitative %ut are qualitative such as gender, race,colourcolour

    or example* holding all other factors constant, female collegeor example* holding all other factors constant, female college

    professors are found to earn less than their male counterpartsprofessors are found to earn less than their male counterparts on whites are found to earn less than whiteson whites are found to earn less than whites  his pattern may result from sex or racial discrimination his pattern may result from sex or racial discrimination $ualitative varia%les such as sex and race does inuence the$ualitative varia%les such as sex and race does inuence the

    dependent varia%le and clearly should %e included among thedependent varia%le and clearly should %e included among the

    explanatory varia%les!explanatory varia%les! $ualitative varia%les usually indicate the presence or a%sence of$ualitative varia%les usually indicate the presence or a%sence of

    a quality such as male or female %lac- or white, literate ora quality such as male or female %lac- or white, literate orilliterate, ur%an or rural, etc the pro%lem is how to incorporateilliterate, ur%an or rural, etc the pro%lem is how to incorporatesuch varia%les in regression along with other quantitativesuch varia%les in regression along with other quantitativevaria%lesvaria%les

  • 8/17/2019 Independent Dummy Variables

    3/32

    33

    $ualitative varia%les in regression$ualitative varia%les in regression

    (&ummy "aria%les)(&ummy "aria%les)

    .ne method of quantifying such attri%utes is %y constructing.ne method of quantifying such attri%utes is %y constructingarti/cial varia%les which ta-e on values of 0 or 1, 1 indicatingarti/cial varia%les which ta-e on values of 0 or 1, 1 indicatingthe a%sence of an attri%ute and 0 indicating the presence ofthe a%sence of an attri%ute and 0 indicating the presence ofthat attri%utethat attri%ute

    "aria%les which assume such 1 and 0 values are called dummy"aria%les which assume such 1 and 0 values are called dummyvaria%les other names for such varia%les are indicatorvaria%les other names for such varia%les are indicatorvaria%les, %inary varia%les qualitative varia%les andvaria%les, %inary varia%les qualitative varia%les andcategorical varia%les!categorical varia%les!

    2e can include dummy varia%les in a regression and conduct2e can include dummy varia%les in a regression and conduct

    hypothesis tests, 3ust as we can any other quantitativehypothesis tests, 3ust as we can any other quantitativevaria%les! 4ut,varia%les! 4ut, interpretation of the coecient on the dummyinterpretation of the coecient on the dummyvariablevariable is somewhat di5erent than what we6ve seen %eforeis somewhat di5erent than what we6ve seen %efore

  • 8/17/2019 Independent Dummy Variables

    4/32

    44

    7onsider a data on annual salary of male7onsider a data on annual salary of male

    and female college teachers and yearsand female college teachers and years

    of teaching experience!of teaching experience!

    &e/ne two dummy varia%les &8 9&e/ne two dummy varia%les &8 9

    &:,one for each category and the model&:,one for each category and the model

    isis

     ; ;ii= annual salary of the ith= annual salary of the ith

    college professorcollege professor

  • 8/17/2019 Independent Dummy Variables

    5/32

    55

    &e/ne one dummy varia%les &8&e/ne one dummy varia%les &8

    for male category and the modelfor male category and the model

    isis

     ;i= annual salary of the ith ;i= annual salary of the ith

    college professor  college professor

  • 8/17/2019 Independent Dummy Variables

    6/32

    66

    'mportant notes'mportant notes

    The assin'ent of " and 1 -alue are ar.itrary

    /ateory that is assined the -alue of " is often referred to as

    control roup0 .ase cateory0 or o'itted cateory in the sense

    that co'parisons are 'ade ith that cateory n this exa'ple

    fe'ale roup

    ntercept ter' α1 in the 'odel is introduced for the control

    roup

  • 8/17/2019 Independent Dummy Variables

    7/32

    77

    'nterpretation of results'nterpretation of results

    ean salary for emale professorean salary for emale professor

    ean salary for male professorean salary for male professor

    ( )   iiii   X  D X Y    β α   +==Ε 

    12 "0( ) ( )

      iiii  X  D X Y    β α α    ++==Ε  212   10

     he estimated equation %y using .S is ;= 8>!> + ?!?> &8+

    >!@0 < (0?!8A)(?!B) (8!C>)

    0) he estimated mean salary of female college teacher is=8>!>+ >!@0<

    8) he estimated mean salary of male college teacher is=:0!1>+ >!@0<

    :) he estimated mean salary of female college teacher with ?years of experience is = 8>!>+>!@0(?) =@0!>?

  • 8/17/2019 Independent Dummy Variables

    8/32

    88

     est of hypothesis est of hypothesis

    21  α+α

    7ompare the annual mean salary of

    male and female college teachers4o5 α26"

    ( )

    ( )ns

    2

    22 7!"7!

    "!$!

    8-ar 

    8t   =

    −=

    α

    α−α=

     ie sex has no effect on 'ean

    salary of collee teachers

    emale

    ale

  • 8/17/2019 Independent Dummy Variables

    9/32

  • 8/17/2019 Independent Dummy Variables

    10/32

    1010

    'ntroduce dummy varia%le for'ntroduce dummy varia%le for

    GE&ERGE&ER

     2here2here

     ;i= 4ase salary of the ith salesperson ;i= 4ase salary of the ith salesperson

  • 8/17/2019 Independent Dummy Variables

    11/32

    1111

    Regressions for dummyRegressions for dummy

    varia%lesvaria%les

    Regression for saleswomen&8=1

    Regression forsalesmen&8=0

  • 8/17/2019 Independent Dummy Variables

    12/32

    1212

    21  α+α

    Saleswomen

    Salesmen

    21  α+α 1α

    Saleswomen

    Salesmen

    1α  1

    α

    Saleswomen

    Salesmen

  • 8/17/2019 Independent Dummy Variables

    13/32

    1313

    SCATTER PLOT

    Salary vs experiene

    E!PER"E#

    403020100

          S      A      L      A      R      $

    14

    13

    12

    11

    10

    9

    8

    7

    6

    %E#&ER

    'e(ale

    (ale

    # l i f h d l# l i f th d l

  • 8/17/2019 Independent Dummy Variables

    14/32

    1414

    #nalysis of the model#nalysis of the model

    #s the regression coeFcient for gender is signi/cant(pvalue H !1@), so we conclude that the /rm doesdiscriminate against its saleswomen! #s coeFcient fordummy varia%le is positive so salesmen6s salary(present category) is more than salesmen6s salary

    4o5 α26" ie 9ender has no influence on salary

  • 8/17/2019 Independent Dummy Variables

    15/32

    1515

    Estimation in regressionEstimation in regression

    Estimated mean salary of Salesman with @ months ofexperience

    Estimated mean salary of Saleswoman with @ months of experience

    :!#7# 

    or #7#!)#(22")"(;7"$!"#8

    =

    =++=iY 

  • 8/17/2019 Independent Dummy Variables

    16/32

    1616

    Regression 2ith wo $ualitative "aria%lesRegression 2ith wo $ualitative "aria%les7onsider the regression of the advertising expenditure7onsider the regression of the advertising expenditure

    on the sales, type of /rm (on the sales, type of /rm (incorporated, notincorporated, not

    incorporatedincorporated) pu%lic relation department and quality of) pu%lic relation department and quality ofsales management (sales management (high, lowhigh, low))

    UDαDααY 33221   +++=

    &2)1 in*rp*ra+e,

      )0 O+-er.ise

    &3)1 i' /ali+y *' sales (anae(en+ -i-

      )0 O+-er.ise

    #*+ in*rp*ra+e, 'ir( .i+- l*. /ali+y

    *' sales (anae(en+ is COPAR"SO#COPAR"SO#

    a+e*ry

  • 8/17/2019 Independent Dummy Variables

    17/32

    1717

    &ata analysis in ''#4&ata analysis in ''#4

    http://c/Documents%20and%20Settings/ARIF/Desktop/example1.MTWhttp://c/Documents%20and%20Settings/ARIF/Desktop/example1.MTWhttp://c/Documents%20and%20Settings/ARIF/Desktop/example1.MTW

  • 8/17/2019 Independent Dummy Variables

    18/32

    1818

    Es+i(a+e, Reressi*n e/a+i*n is

    $)4400016000&25500&3 "#TERPRETAT"O# Averae expen,i+re '*r ase a+e*ry

    n*+ in*p*ra+e, .i+- l*. (anae(en+

      is 44000

    T-e expen,i+re is ,erease y 16000 i' +-e 'ir( is in*rp*ra+e,Expen,i+re is inrease y 5500

     i' /ali+y *' sale (anae(en+ is -i-:

    #*nsini'ian+ vale +ra+i* '*r &2 in,ia+es +-a+ +-ere is n* ,i''erene e+.een +-e

     A,ver+isin expen,i+re *' in*rp*ra+e, an, n*+ in*rp*ra+e, 'ir(s

    #*nsini'ian+ vale +ra+i* '*r &3 in,ia+es +-a+ +-ere is n* ,i''erene e+.een +-e

     A,ver+isin expen,i+re '*r +-e 'ir(s -avin -i- *r l*. /ali+y sales (anae(en+

    &8=0 incorporated&8=0 incorporated

      =1 .therwise(i!e not incorporated)  =1 .therwise(i!e not incorporated)

    (&:=0 if quality of sales management high(&:=0 if quality of sales management high

      =1 .therwise( ow sales management)  =1 .therwise( ow sales management)

  • 8/17/2019 Independent Dummy Variables

    19/32

    1919

    Regression 2ith more than two categoriesSuppose that %ased on the cross sectional data we want toregress the annual expenditure on healthcare %y anindividual on the income and education of the individual!

    Since the varia%le education is qualitative in nature weconsider three mutually exclusive levels of education (i)ess than high school (ii) Iigh School (iii) 7ollege 

    UDαDααY 33221   +++=

     ;i= #nnual expenditure on health care

  • 8/17/2019 Independent Dummy Variables

    20/32

  • 8/17/2019 Independent Dummy Variables

    21/32

    2121

    'nterpretation of results'nterpretation of results

    ean healthcare expenditure %y an individual having less than high school e

     X  D DY  E    β α   +=== 132 )"0"

  • 8/17/2019 Independent Dummy Variables

    22/32

    2222

    TET !" H#P!THE$TET !" H#P!THE$0) 7omparison of mean expenditure on health care %y an0) 7omparison of mean expenditure on health care %y an

    individual having high school education and less than highindividual having high school education and less than high

    school education ( 4ase group)school education ( 4ase group)

    8) 7omparison of mean expenditure on health care %y an8) 7omparison of mean expenditure on health care %y anindividual having college education and less than high schoolindividual having college education and less than high school

    education ( 4ase group)education ( 4ase group)

    "α542"

      =

    "α54 3"   =

    Expendie= -4.41 + 7.92D2 +10.5 D3 +0.361X

    SE   (3.195) (3.328) (7.813) (0.122)

    t-ratio   2.38* 1.34 (2.95)

    P-Value   0.037 0.206 0.013

    TET !" H#P!THE$TET !" H#P!THE$

  • 8/17/2019 Independent Dummy Variables

    23/32

    2323

    TET !" H#P!THE$TET !" H#P!THE$ :) 7omparison of mean expenditure on health care %y an individual having high school:) 7omparison of mean expenditure on health care %y an individual having high school

    education and collegeeducation and college

    32" αα54   =

    Unrestricted Model *estricted %odel 

    i3i32i21 UDαDααY   ++++=   X β    32" αα54   =  32

    UDααY

    =

    2

    i2i=

    21

     D D DWhere

     X 

    +=

    +++=   β  

    Unrestricted ANOVA Restricted ANOVA

    SOV DF SS

    Regression

    Error

    Total

    3

    11

    14

    3!"#$

    %1#1

    31"4#

    SOV DF SS

    Regression

    Error

    Total

    "

    1"

    14

    3!1#"

    %"#&

    31"4#

    ( ) ( )   ( ) ( )   ns

    URUR

    UR RUR R

     Edf   ESS 

     Edf   Edf   ESS  ESS  F    2!1#"

    #!

    "1

    11

  • 8/17/2019 Independent Dummy Variables

    24/32

    2424

    &K; "#R'#4E .R 7.L#R'G&K; "#R'#4E .R 7.L#R'G

     2. REGRESS'. 'ES 2. REGRESS'. 'ES

    2hen we use a regression model involving time series data, it2hen we use a regression model involving time series data, it

    may happen that there is amay happen that there is a structural changestructural change in thein the

    relationship %etween the regressand and the regressorsrelationship %etween the regressand and the regressors 

    4y structural change we mean that the value of the parameters4y structural change we mean that the value of the parametersof the model do not remain the same through the entire timeof the model do not remain the same through the entire time

    periodperiod!! &ummy varia%le can %e used to /nd structural&ummy varia%le can %e used to /nd structural

    changes in the regression parameters!changes in the regression parameters!

    7oncurrent Larallel &issimilar Equal or

    7oincident

    E%uality of two regressionE

    %uality of two regression

  • 8/17/2019 Independent Dummy Variables

    25/32

    2525

    E%uality of two regressionE%uality of two regression

    lines by dummy variablelines by dummy variable

    approach:approach:E

  • 8/17/2019 Independent Dummy Variables

    26/32

    2626

    ( ) ( )β βα α UD X D X2i i 2i i1 2 i1 2ii y   =   + + + + − − − −

     ;i = saving of a family

  • 8/17/2019 Independent Dummy Variables

    27/32

    2727

     he equality of two saving function can %e he equality of two saving function can %e

    tested %y considering the hypothesistested %y considering the hypothesis

  • 8/17/2019 Independent Dummy Variables

    28/32

    2828

    S."S." && SSSS

    RegRegErrorError

    ::0?0?

    0?C>?C:>0?C>?C:>?:8@:8???:8@:8??

     otal otal 0B0B  A1111111A1111111

    S."S." && SSSS

    RegReg

    ErrorError

    00

    0A0A

    0@A1?@?00@A1?@?0

    ?>0B:>:B?>0B:>:B

     otal otal 0B0B A1111111A1111111

    Unrestricted %odel

    ( )β βα α UD X D X2i i 2i i1 2 i1 2i

     y   =   + + + +*estricted%odel

    iβα   Xi1   1i U  y  =   + +

    ( ) ( ) ( ) ( )< !$173$37 !32#32!! < 1; 1!< !32#32!! < 1!

    $("";!#"11

    37#332712#

     R UR R UR

    UR UR

    ns

     ESS ESS Edf Edf   F  ESS Edf  

    − − − −= =

    = =

    Regression ines for Kr%an 9 Rural are

    E

  • 8/17/2019 Independent Dummy Variables

    29/32

    2929

    E

  • 8/17/2019 Independent Dummy Variables

    30/32

    3030

    'ntroduce dummy varia%le for'ntroduce dummy varia%le for

    structural changestructural change

     ;i= saving

  • 8/17/2019 Independent Dummy Variables

    31/32

    3131

    Scatter plotScatter plot

  • 8/17/2019 Independent Dummy Variables

    32/32

    3232

    S."S." && SSSS

    RegReg

    ErrorError

    ::

    8888

    AA1CB!A:>AA1CB!A:>

    00CB1!8@:00CB1!8@:

     otal otal 8@8@ BBAC1!1ACBBAC1!1AC

    S."S." && SSSS

    RegRegErrorError

    008>8>

    C??80!CAAC??80!CAA8:8>A!8BA8:8>A!8BA

     otal otal 8@8@ BBAC1!1ACBBAC1!1AC

    *estricted%odel

    iβα   Xi1   1

    iU  y   =   + +

    Regression are di5erent for two time

    ( ) ( )( )

    =!71"72#3#

    "23#27

    22

    2#3117"

    )222$(

    2#3117"27;232$;

    <

    <==

    =−−

    =

    URUR

    UR RUR R

     Edf   ESS 

     Edf   Edf   ESS  ESS  F 

    )("!#"";"$71#2"1!18iii

      DX  X  DY    −++=i X Y  "3;"$23!2

    8 +=

    iiiii   U  X  D X  DY    +−++=   )(

    'odel*estricted>Un

    1121   β β α α 


Recommended