test on variables in statistics

Upload: nirmal-modh

Post on 10-Apr-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Test on Variables in Statistics

    1/24

    Test On VariablesTest On Variables

    Group 5Group 5

    AsimAsim KumarKumar VermaVermaDeeptimanDeeptiman GuhaGuhaHimanshuHimanshu AroraAroraKaavishKaavish KidwaiKidwaiNirmalNirmal ModhModh

    In Surveys, the foolish ask questions, wise cannot answersIn Surveys, the foolish ask questions, wise cannot answers..--Oscar WildeOscar Wilde

  • 8/8/2019 Test on Variables in Statistics

    2/24

  • 8/8/2019 Test on Variables in Statistics

    3/24

    Two Types ofVariablesTwo Types ofVariables

    Quantitative:Quantitative: Numericvaluethatitmakessenseto do arithmeticNumericvaluethatitmakessenseto do arithmetic

    operations (+,operations (+, --, x, /), x, /) Findinganarithmetic average makessenseFindinganarithmetic average makessense

    EX:EX:

    HeightHeight

    WeightWeight AgeAge

    IncomeIncome

    Test ScoresTest Scores

  • 8/8/2019 Test on Variables in Statistics

    4/24

    Types ofVariablesTypes ofVariables

    CategoricalCategorical

    Records which ofseveralgroups or categoriesRecords which ofseveralgroups or categoriesto whichanindividualbelongsto whichanindividualbelongs

    Ifthereare only 2 possiblecategories,theIfthereare only 2 possiblecategories,thevariableiscalledbinaryvariableiscalledbinary

    Ex:Ex: GenderGender

    RaceRace

    Marital statusMarital status

  • 8/8/2019 Test on Variables in Statistics

    5/24

    Stepsin Hypothesis TestingStepsin Hypothesis Testing

    1.1. Hypothesis formulationHypothesis formulation

    2.2. SpecifythesignificancelevelSpecifythesignificancelevel

    3.3. CollectthedataandcalculatetheteststatisticCollectthedataandcalculatetheteststatistic4.4. FetchthetabulatedvalueFetchthetabulatedvalue

    5.5. MakethestatisticaldecisionMakethestatisticaldecision

  • 8/8/2019 Test on Variables in Statistics

    6/24

    Statistical Test

    Parametric Non Parametric

    When ngreater than 30

    When nless than

    30

  • 8/8/2019 Test on Variables in Statistics

    7/24

    HypothesistestingHypothesistesting

    Goal:Makestatement(s) regarding unknown populationGoal:Makestatement(s) regarding unknown populationparameter valuesbased onsampledataparameter valuesbased onsampledata

    Elements ofahypothesistest:Elements ofahypothesistest: NullhypothesisNullhypothesis -- Statementregardingthevalue(s) ofStatementregardingthevalue(s) of

    unknown parameter(s). Typically willimplyno associationunknown parameter(s). Typically willimplyno associationbetween explanatoryand responsevariablesin our applicationsbetween explanatoryand responsevariablesin our applications

    Alternative hypothesisAlternative hypothesis-- Statementcontradictoryto thenullStatementcontradictoryto thenullhypothesis (willalwayscontainaninequality)hypothesis (willalwayscontainaninequality)

    TeststatisticTeststatistic-- Quantitybased onsampledataandnullQuantitybased onsampledataandnull

    hypothesis usedto testbetweennullandalternativehypothesis usedto testbetweennullandalternativehypothesishypothesis RejectionregionRejectionregion-- Values oftheteststatistic for whichweValues oftheteststatistic for whichwe

    rejectthenullin favor ofthealternativehypothesisrejectthenullin favor ofthealternativehypothesis

  • 8/8/2019 Test on Variables in Statistics

    8/24

    TestOnVariablesTestOnVariables

    Inthesedecisionsaremade on quantitativemeasuresInthesedecisionsaremade on quantitativemeasures

    T testT test

    StatisticalStatistical hypothesishypothesis testtest inin whichwhich thethetesttest statisticstatisticfollowsfollows aa student'sstudent's t t dist ributiondistribution ifif thethe nullnullhypothesishypothesis isis truetruenn > 3030

  • 8/8/2019 Test on Variables in Statistics

    9/24

    1.1. StateStateHH00 HH00:: QuQu3.03.0

    2.2. StateStateHH11 HH11 ::QQ

    3.3. ChooseChoose EE EE = .05= .05

    4.4. ChooseChoose nn n= 100n= 100

    5.5. Choose TestChoose Test:: Z testZ test

    Hypothesis Testing: Steps

    Test the Assumption that the true meangrade point average of juniors is at least 3.

  • 8/8/2019 Test on Variables in Statistics

    10/24

    6. Set Up Critical Value(s)6. Set Up Critical Value(s) Ze= 1.645Ze= 1.645

    7. Collect Data7. Collect Data 100 students sampled100 students sampled

    8. Compute Test Statistic8. Compute Test Statistic Computed Test Stat.=Computed Test Stat.= --22

    9. Make Statistical Decision9. Make Statistical Decision Since Z mod greater than ZeSince Z mod greater than ZeReject NullHypothesisReject NullHypothesis

    10. Express Decision10. Express Decision The true mean grade point isThe true mean grade point isless than 3.0less than 3.0

    Hypothesis Testing: Steps

    Test the Assumption that grade point average ofjuniors is at least 3.

    (continued)

  • 8/8/2019 Test on Variables in Statistics

    11/24

    Z0

    E

    Reject H0

    Z0

    Reject H0

    E

    H0: Qu

    H1: Q < 0H

    0: Qe0

    H1: Q > 0

    Must BeSignificantlyBelowQ=0

    Small values dont contradictH0

    Dont Reject H0!

    RejectionRegion

  • 8/8/2019 Test on Variables in Statistics

    12/24

    Hypothesis TestHypothesis Test HypothesisHypothesis testingtesting isis usedused whenwhen wewe wantwant toto

    knowknow whetherwhether thethe samplesample datadata supportsupport somesomepreconceived preconceived theorytheory wewe holdhold aboutabout thethe

    parametersparameters ofof thethe populationpopulation modelmodel.. IfIf wewe cancan findfind aa numericalnumerical valuevalue ofof thethe

    relevantrelevant probability,probability, thenthen wewe mightmight bebe ableable totoassessassess whetherwhether thethe beliefbelief isis supportedsupported oror notnot::

    aa highhigh probability probability ofof obtainingobtaining thethe datadatasupportssupports thethe belief,belief, aa lowlow oneone doesdoes notnot..

  • 8/8/2019 Test on Variables in Statistics

    13/24

    ttestttest

    tt teststests isis simplysimply comparingcomparing twotwo meansmeans toto seesee ifif theythey

    areare significantlysignificantly differentdifferent fromfrom eacheach otherother..

    TheThe moremore technicaltechnical definitiondefinition oror descriptiondescription ofof aa tt testtestisis anyany statisticalstatistical testtest thatthat usesuses thethe t,t, oror Student'sStudent's t,t,

    familyfamily ofof distributionsdistributions..

    MostMost EconomistsEconomists thinkthink ofof godgod asas workingworking greatgreat multiplemultiple

    regressionsregressions inin thethe skysky EdgarEdgar FiedlerFiedler

  • 8/8/2019 Test on Variables in Statistics

    14/24

    StudentstStudentst--testtestIfIf randomrandom samplessamples ofof sizesize lessless thanthan 3030 areare takentaken fromfrom aa normalnormal distributiondistribution andand thethe samplessamples

    usedused toto estimateestimate thethe variance,variance, thenthen thethe statisticstatistic

    --

    s/ns/n

    isis notnot normallynormally distributeddistributed.. TheThe probabilitiesprobabilities inin thethe tailstails ofof thisthis distributiondistribution areare greatergreaterthanthan forfor thethe standardstandard normalnormal distributiondistribution

    ThisThis isis reasonablereasonable sincesince

    --

    z=z=

    s/ns/n

    containscontains onlyonly oneone randomrandom variablevariable ,, whilewhile--

    s/ns/n

    containscontains twotwo randomrandom variablesvariables yy andand ss.. AsAs nn increasesincreases thisthis newnew distributiondistribution approachesapproachesthethe standardstandard normalnormal distributiondistribution..

  • 8/8/2019 Test on Variables in Statistics

    15/24

    Properties of t TestProperties of t Test

    StudentsStudents tt distributionsdistributions areare

    11.. UnimodalUnimodal;;

    22.. AsymptoticAsymptotic toto thethe horizontalhorizontal axisaxis;;33.. SymmetricalSymmetrical aboutabout zero,zero, E(t)E(t);;

    44.. DependentDependent onon v,v, thethe degreesdegrees ofof freedomfreedom (for(for thethestatisticstatistic underunder discussion,discussion, v=nv=n--11));;

    55.. MoreMore variablevariable thanthan thethe standardstandard normalnormaldistribution,distribution, V(t)=v/(vV(t)=v/(v--22)) forfor nn >> 22;;

    66.. ApproximatelyApproximately standardstandard normalnormal ifif vv isis largelarge..

  • 8/8/2019 Test on Variables in Statistics

    16/24

    Excel CommandExcel Command

    ToTo testtest whetherwhether thethe populationpopulation meanmean couldcould bebe ,,useuse ::

    MeanTestMeanTest usesuses thethe originaloriginal datadata andand thethehypotheticalhypothetical valuevalue ofof thethe populationpopulation meanmean..

    StudentTPValueStudentTPValue usesuses onlyonly thethe valuevalue ofof thethe testteststatisticstatistic tt == mm--//SmSm andand degreesdegrees ofof freedomfreedom nn -- 11..

    TheThe twotwo commandscommands useuse thethe StudentStudent tt distributiondistributiontoto performperform thethe testingtesting..

  • 8/8/2019 Test on Variables in Statistics

    17/24

    ComparisionComparision ofstandardnormaldistributionandatofstandardnormaldistributionandatdistributiondistribution

  • 8/8/2019 Test on Variables in Statistics

    18/24

    StandardErrorStandardError

    TestTest determinesdetermines whetherwhether differencesdifferences inin twotwo

    samplessamples isis significantsignificant enoughenough toto suggestsuggest aadifferencedifference inin respectiverespective populationspopulations..

    HenceHence averageaverage expectedexpected differencedifference betweenbetween thethemeansmeans ofof twotwo samplessamples isis requiredrequired.. ThisThis averageaveragedifferencedifference isis knownknown asas thethe StandardStandard ErrorError ofof thethedifferencedifference betweenbetween twotwo meansmeans..

  • 8/8/2019 Test on Variables in Statistics

    19/24

    Paired SamplesPaired Samples

    AA dependentdependent samplessamples tt testtest isis alsoalso usedused toto comparecompare twotwomeansmeans onon aa singlesingle dependentdependent variablevariable..

    UnlikeUnlike thethe independentindependent samplessamples test,test, however,however, aa dependentdependent

    samplessamples tt testtest isis usedused toto comparecompare thethe meansmeans ofof aa singlesinglesamplesample oror ofof twotwo matchedmatched oror pairedpaired samplessamples..

    ForFor example,example, ifif aa groupgroup ofof studentsstudents tooktook aa mathmath testtest ininMarchMarch andand thatthat samesame groupgroup ofof studentsstudents tooktook thethe samesame mathmathtesttest twotwo monthsmonths laterlater inin May,May, wewe couldcould comparecompare theirtheir

    averageaverage scoresscores onon thethe twotwo testtest datesdates usingusing aa dependentdependentsamplessamples tttesttest..

  • 8/8/2019 Test on Variables in Statistics

    20/24

    ExampleExample

    SupposeSuppose wewe wantwant toto knowknow whetherwhether employeesemployees atat ourourwidgetwidget--makingmaking factoryfactory areare moremore productiveproductive afterafter theythey returnreturnfromfrom aa 22--weekweek vacationvacation..

    WeWe randomlyrandomly selectselect 3030 ofof mymy employeesemployees andand calculatecalculate thetheaverageaverage numbernumber ofof widgetswidgets mademade byby eacheach employeeemployee duringduringthethe weekweek beforebefore theythey gogo onon vacationvacation..

    WeWe findfind that,that, onon average,average, thethe employeesemployees mademade 250250 widgetswidgetseacheach duringduring thethe weekweek.. DuringDuring thethe weekweek afterafter theythey returnreturnfromfrom vacation,vacation, II keepkeep tracktrack ofof howhow manymany widgetswidgets isis mademade

    by by thethe samesame samplesample ofof 3030 employeesemployees andand findfind that,that, ononaverage,average, theythey mademade 300300 widgetswidgets eacheach duringduring thethe weekweek afterafterreturningreturning fromfrom theirtheir vacationsvacations..

  • 8/8/2019 Test on Variables in Statistics

    21/24

  • 8/8/2019 Test on Variables in Statistics

    22/24

    ContdContd

    T=observedT=observed diffdiff.. betweenbetween prevacationprevacation andand postvacationpostvacationmeans/standardmeans/standard errorerror ofof thethe differencedifference betweenbetween thethe meansmeans

    OrOr

    T=xT=x--y/y/SdSd

    TheThe formulaformula forfor calculatingcalculating thethe standardstandard errorerror ofof thethedifferencedifference between between thethe meansmeans forfor dependentdependent samplessamples isis

    slightlyslightly differentdifferent thanthan thethe oneone forfor independentindependent samples,samples, butbutthethe principles principles involvedinvolved (i(i..ee..,, whatwhat thethe standardstandard errorerrorrepresents)represents) areare thethe samesame..

  • 8/8/2019 Test on Variables in Statistics

    23/24

    CaseCase

    Effectiveness ofNewly ImplementedEffectiveness ofNewly ImplementedGreen BeltProject ina Call Center ofGreen BeltProject ina Call Center of

    Insurance CompanyInsurance Company

  • 8/8/2019 Test on Variables in Statistics

    24/24

    BibliographyBibliography ParadigmParadigm : IMT G Journal: IMT G Journal

    Appliedstatisticsinbusinessand economicsbyAppliedstatisticsinbusinessand economicsbyDoaneDoane & Seward& Seward

    www.w

    ikipedia.

    comwww.w

    ikipedia.

    com Case: An analysis of the Indian Textile IndustryCase: An analysis of the Indian Textile Industry

    in Quota Free Regime byin Quota Free Regime by ManishaManisha Sharma &Sharma &AnuAnu PrashaantPrashaant

    It is not nice to be wedded to anythingIt is not nice to be wedded to anything not even anot even atheorytheory Samuel ButlerSamuel Butler