ISYE6414 (Regression) Midterm 2
Questions And Answers
What iis icooks idistance iused ifor? i- icorrect iAnswers i✔✔ i-It imeasures ihow imuch iall iof ithe ivalues iin
ithe iregression imodel ichange iwith ithe iith iobservation iis iremoved. iBasically iits ia itest ifor ioutliers
Rule iof ithumb: iD idenotes icooks idistance, iif iD iis i> i4/n
OR iD i> i1 ior iany ilarge iD ithen iit imay ibe ian ioutlier iand ishould ibe iremoved.
If ithe inormality iassumption idoes inot ihold, iwe ican ipursue ia itransformation iin ithe iresponse ivariable.
iT/F i- icorrect iAnswers i✔✔ i-True
If ithe ilinearity iassumption idoes inot ihold, iwe ican ipursue ia itransformation iin ithe iresponse ivariable.
iT/F i- icorrect iAnswers i✔✔ i-False, iwe ipursue ia itransformation iin ithe ipredictor ivariables.
R^2 iwill ialways iincrease iif iwe iadd imore ipredicting ivariables. iT/F i- icorrect iAnswers i✔✔ i-True
If iwe iwant ito icompare imodels iwith idifferent inumbers iof ipredicting ivariables, iwhat istatistic ishould
iwe iuse? i- icorrect iAnswers i✔✔ i-Adjusted iR^2 ibecause iit iadjusts ifor ithe inumber iof ipredicting
ivariables. iIt idoesn't iincrease iwhen iwe iadd imore ipredicting ivariables.
A istatistic ithat ieffectively isummarizes ihow iwell ithe iX's iare ilinearly irelated ito iY iis ithe icorrelation
icoefficient. iT/F i- icorrect iAnswers i✔✔ i-True
T/F i- iThe icorrelation icoefficient icannot ibe iused ito ievaluate ithe icorrelation ibetween ithe ipredicting
ivariables ifor idetecting i(near) ilinear idependence iamong ithe ivariables i(or imulticolinearity) i- icorrect
iAnswers i✔✔ i-False, iit iCAN
How ido iyou idiagnose imulticolinearity? i- icorrect iAnswers i✔✔ i-Calculate ithe iVIF i(variance iinflation
ifactor) ifor ieach ipredicting ivariable
VIF i= i1 i/ i(1 i- iR^2j)
, If iVIF i< imax(10, i1 i/ i(1- iR^2)) ithen iwe igot ia iproblem
If ia ivariable iis icorrelated ibut idoes inot ihave imulticolinearity, iis ithis ia iproblem? i- icorrect iAnswers
i✔✔ i-Not inecessarily ibruh
What idoes ithe iVIF imeasure i- icorrect iAnswers i✔✔ i-the iVIF imeasures ithe iproportional iincrease iin
ithe ivariance iof ibeta ihat icompared ito iwhat iit iwould ihave ibeen iif ithe ipredicting ivariables ihad ibeen
icompletely iuncorrelated.
True/False: iThe iresponse ivariable iin ilogistic iregression iis ia ibinary iresponse? i- icorrect iAnswers i✔✔ i-
True
True/False: iIn ilogistic iregression, iwe imodel ithe iprobability iof ia isuccess igiven ithe ipredicting
ivariables, inot ithe iresponse iitself. i- icorrect iAnswers i✔✔ i-True
What iare ithe iassumptions ifor ilogistic iregression? i- icorrect iAnswers i✔✔ i-Linearity iAssumption
Independence iAssumption
The iG-Link ifunction iis ia ilogit ifunction iAssumption
What iis ithe ilogit ifunction? i- icorrect iAnswers i✔✔ i-ratio ibetween ithe iprobability iof isuccess iover
iprobability iof ia ifailure. iSo ibasically iratio ibetween ilog iof iP iover i1-p
What iis ithe iinterpretation iof ithe ilogistic iregression icoefficient? i- icorrect iAnswers i✔✔ i-The ilog iof
ithe iodds iratio ifor ian iincrease iof ione iunit iin ithe ipredicting ivariable. iWe ido inot iinterpret ibeta iwith
irespect ito ithe iresponse ivariable ibut iwith irespect ito ithe iodds iof isuccess.
How imany iregression icoefficients iare ithere ifor ilogistic iregression? i- icorrect iAnswers i✔✔ i-Since
ithere iis ino ierror itime, iyou ihave iP i+ i1 iwith iintercept.
Logistic iregression iis idifferent ifrom istandard ilinear iregression iin ithat
a) iIt idoes inot ihave ian ierror iterm
b) iThe iresponse ivariable iis inot inormally idistributed.
c) iIt imodels iprobability iof ia iresponse iand inot ithe iexpectation iof ithe iresponse.
d) iAll iof ithe iabove. i- icorrect iAnswers i✔✔ i-d) iall iof ithe iabove
Which ione iis icorrect?
a) iThe ilogit ilink ifunction iis ithe ionly ilink ifunction ithat ican ibe iused ifor imodeling ibinary iresponse
idata.
b) iLogistic iregression imodels ithe iprobability iof ia isuccess igiven ia iset iof ipredicting ivariables.
Questions And Answers
What iis icooks idistance iused ifor? i- icorrect iAnswers i✔✔ i-It imeasures ihow imuch iall iof ithe ivalues iin
ithe iregression imodel ichange iwith ithe iith iobservation iis iremoved. iBasically iits ia itest ifor ioutliers
Rule iof ithumb: iD idenotes icooks idistance, iif iD iis i> i4/n
OR iD i> i1 ior iany ilarge iD ithen iit imay ibe ian ioutlier iand ishould ibe iremoved.
If ithe inormality iassumption idoes inot ihold, iwe ican ipursue ia itransformation iin ithe iresponse ivariable.
iT/F i- icorrect iAnswers i✔✔ i-True
If ithe ilinearity iassumption idoes inot ihold, iwe ican ipursue ia itransformation iin ithe iresponse ivariable.
iT/F i- icorrect iAnswers i✔✔ i-False, iwe ipursue ia itransformation iin ithe ipredictor ivariables.
R^2 iwill ialways iincrease iif iwe iadd imore ipredicting ivariables. iT/F i- icorrect iAnswers i✔✔ i-True
If iwe iwant ito icompare imodels iwith idifferent inumbers iof ipredicting ivariables, iwhat istatistic ishould
iwe iuse? i- icorrect iAnswers i✔✔ i-Adjusted iR^2 ibecause iit iadjusts ifor ithe inumber iof ipredicting
ivariables. iIt idoesn't iincrease iwhen iwe iadd imore ipredicting ivariables.
A istatistic ithat ieffectively isummarizes ihow iwell ithe iX's iare ilinearly irelated ito iY iis ithe icorrelation
icoefficient. iT/F i- icorrect iAnswers i✔✔ i-True
T/F i- iThe icorrelation icoefficient icannot ibe iused ito ievaluate ithe icorrelation ibetween ithe ipredicting
ivariables ifor idetecting i(near) ilinear idependence iamong ithe ivariables i(or imulticolinearity) i- icorrect
iAnswers i✔✔ i-False, iit iCAN
How ido iyou idiagnose imulticolinearity? i- icorrect iAnswers i✔✔ i-Calculate ithe iVIF i(variance iinflation
ifactor) ifor ieach ipredicting ivariable
VIF i= i1 i/ i(1 i- iR^2j)
, If iVIF i< imax(10, i1 i/ i(1- iR^2)) ithen iwe igot ia iproblem
If ia ivariable iis icorrelated ibut idoes inot ihave imulticolinearity, iis ithis ia iproblem? i- icorrect iAnswers
i✔✔ i-Not inecessarily ibruh
What idoes ithe iVIF imeasure i- icorrect iAnswers i✔✔ i-the iVIF imeasures ithe iproportional iincrease iin
ithe ivariance iof ibeta ihat icompared ito iwhat iit iwould ihave ibeen iif ithe ipredicting ivariables ihad ibeen
icompletely iuncorrelated.
True/False: iThe iresponse ivariable iin ilogistic iregression iis ia ibinary iresponse? i- icorrect iAnswers i✔✔ i-
True
True/False: iIn ilogistic iregression, iwe imodel ithe iprobability iof ia isuccess igiven ithe ipredicting
ivariables, inot ithe iresponse iitself. i- icorrect iAnswers i✔✔ i-True
What iare ithe iassumptions ifor ilogistic iregression? i- icorrect iAnswers i✔✔ i-Linearity iAssumption
Independence iAssumption
The iG-Link ifunction iis ia ilogit ifunction iAssumption
What iis ithe ilogit ifunction? i- icorrect iAnswers i✔✔ i-ratio ibetween ithe iprobability iof isuccess iover
iprobability iof ia ifailure. iSo ibasically iratio ibetween ilog iof iP iover i1-p
What iis ithe iinterpretation iof ithe ilogistic iregression icoefficient? i- icorrect iAnswers i✔✔ i-The ilog iof
ithe iodds iratio ifor ian iincrease iof ione iunit iin ithe ipredicting ivariable. iWe ido inot iinterpret ibeta iwith
irespect ito ithe iresponse ivariable ibut iwith irespect ito ithe iodds iof isuccess.
How imany iregression icoefficients iare ithere ifor ilogistic iregression? i- icorrect iAnswers i✔✔ i-Since
ithere iis ino ierror itime, iyou ihave iP i+ i1 iwith iintercept.
Logistic iregression iis idifferent ifrom istandard ilinear iregression iin ithat
a) iIt idoes inot ihave ian ierror iterm
b) iThe iresponse ivariable iis inot inormally idistributed.
c) iIt imodels iprobability iof ia iresponse iand inot ithe iexpectation iof ithe iresponse.
d) iAll iof ithe iabove. i- icorrect iAnswers i✔✔ i-d) iall iof ithe iabove
Which ione iis icorrect?
a) iThe ilogit ilink ifunction iis ithe ionly ilink ifunction ithat ican ibe iused ifor imodeling ibinary iresponse
idata.
b) iLogistic iregression imodels ithe iprobability iof ia isuccess igiven ia iset iof ipredicting ivariables.
