Mathematicalstatistics
FromWikipedia,thefreeencyclopedia
Mathematicalstatisticsistheapplicationof
mathematicstostatistics,whichwasoriginally
conceivedasthescienceofthestatethecollection
andanalysisoffactsaboutacountry:itseconomy,land,
military,population,andsoforth.Mathematical
techniqueswhichareusedforthisincludemathematical
analysis,linearalgebra,stochasticanalysis,differential
equations,andmeasuretheoreticprobabilitytheory.[1][2]
Contents
1 Introduction
2 Topics
2.1 Probabilitydistributions
2.1.1 Specialdistributions
2.2 Statisticalinferences
2.3 Regression
2.4 Nonparametricstatistics
3 Statistics,mathematics,andmathematical
statistics
4 Seealso
5 References
6 Additionalreading
Illustrationoflinearregressiononadataset.
Regressionanalysisisanimportantpartof
mathematicalstatistics.
Introduction
Statisticalscienceisconcernedwiththeplanningofstudies,especiallywiththedesignofrandomized
experimentsandwiththeplanningofsurveysusingrandomsampling.Theinitialanalysisofthedatafrom
properlyrandomizedstudiesoftenfollowsthestudyprotocol.
Ofcourse,thedatafromarandomizedstudycanbeanalyzedtoconsidersecondaryhypothesesorto
suggestnewideas.Asecondaryanalysisofthedatafromaplannedstudyusestoolsfromdataanalysis.
Dataanalysisisdividedinto:
descriptivestatisticsthepartofstatisticsthatdescribesdata,i.e.summarisesthedataandtheir
typicalproperties.
inferentialstatisticsthepartofstatisticsthatdrawsconclusionsfromdata(usingsomemodelforthe
data):Forexample,inferentialstatisticsinvolvesselectingamodelforthedata,checkingwhetherthe
datafulfilltheconditionsofaparticularmodel,andwithquantifyingtheinvolveduncertainty(e.g.
usingconfidenceintervals).
Whilethetoolsofdataanalysisworkbestondatafromrandomizedstudies,theyarealsoappliedtoother
kindsofdataforexample,fromnaturalexperimentsandobservationalstudies,inwhichcasethe
inferenceisdependentonthemodelchosenbythestatistician,andsosubjective.[3]
Mathematicalstatisticshasbeeninspiredbyandhasextendedmanyoptionsinappliedstatistics.
Topics
Thefollowingaresomeoftheimportanttopicsinmathematicalstatistics:[4][5]
Probabilitydistributions
Aprobabilitydistributionassignsaprobabilitytoeachmeasurablesubsetofthepossibleoutcomesofa
randomexperiment,survey,orprocedureofstatisticalinference.Examplesarefoundinexperimentswhose
samplespaceisnonnumerical,wherethedistributionwouldbeacategoricaldistributionexperiments
whosesamplespaceisencodedbydiscreterandomvariables,wherethedistributioncanbespecifiedbya
probabilitymassfunctionandexperimentswithsamplespacesencodedbycontinuousrandomvariables,
wherethedistributioncanbespecifiedbyaprobabilitydensityfunction.Morecomplexexperiments,such
asthoseinvolvingstochasticprocessesdefinedincontinuoustime,maydemandtheuseofmoregeneral
probabilitymeasures.
Aprobabilitydistributioncaneitherbeunivariateormultivariate.Aunivariatedistributiongivesthe
probabilitiesofasinglerandomvariabletakingonvariousalternativevaluesamultivariatedistribution(a
jointprobabilitydistribution)givestheprobabilitiesofarandomvectorasetoftwoormorerandom
variablestakingonvariouscombinationsofvalues.Importantandcommonlyencounteredunivariate
probabilitydistributionsincludethebinomialdistribution,thehypergeometricdistribution,andthenormal
distribution.Themultivariatenormaldistributionisacommonlyencounteredmultivariatedistribution.
Specialdistributions
Normaldistribution(Gaussiandistribution),themostcommoncontinuousdistribution
Bernoullidistribution,fortheoutcomeofasingleBernoullitrial(e.g.success/failure,yes/no)
Binomialdistribution,forthenumberof"positiveoccurrences"(e.g.successes,yesvotes,etc.)given
afixedtotalnumberofindependentoccurrences
Negativebinomialdistribution,forbinomialtypeobservationsbutwherethequantityofinterestis
thenumberoffailuresbeforeagivennumberofsuccessesoccurs
Geometricdistribution,forbinomialtypeobservationsbutwherethequantityofinterestisthe
numberoffailuresbeforethefirstsuccessaspecialc*Discreteuniformdistribution,forafinitesetof
values(e.g.theoutcomeofafairdie)
Continuousuniformdistribution,forcontinuouslydistributedvalues
Poissondistribution,forthenumberofoccurrencesofaPoissontypeeventinagivenperiodoftime
Exponentialdistribution,forthetimebeforethenextPoissontypeeventoccurs
Gammadistribution,forthetimebeforethenextkPoissontypeeventsoccur
Chisquareddistribution,thedistributionofasumofsquaredstandardnormalvariablesusefule.g.
forinferenceregardingthesamplevarianceofnormallydistributedsamples(seechisquaredtest)
Student'stdistribution,thedistributionoftheratioofastandardnormalvariableandthesquareroot
ofascaledchisquaredvariableusefulforinferenceregardingthemeanofnormallydistributed
sampleswithunknownvariance(seeStudent'sttest)
Betadistribution,forasingleprobability(realnumberbetween0and1)conjugatetotheBernoulli
distributionandbinomialdistribution
Statisticalinferences
Statisticalinferenceistheprocessofdrawingconclusionsfromdatathataresubjecttorandomvariation,
forexample,observationalerrorsorsamplingvariation.[6]Initialrequirementsofsuchasystemof
proceduresforinferenceandinductionarethatthesystemshouldproducereasonableanswerswhenapplied
towelldefinedsituationsandthatitshouldbegeneralenoughtobeappliedacrossarangeofsituations.
Inferentialstatisticsareusedtotesthypothesesandmakeestimationsusingsampledata.Whereas
descriptivestatisticsdescribeasample,inferentialstatisticsinferpredictionsaboutalargerpopulationthat
thesamplerepresents.
Theoutcomeofstatisticalinferencemaybeananswertothequestion"whatshouldbedonenext?",where
thismightbeadecisionaboutmakingfurtherexperimentsorsurveys,oraboutdrawingaconclusionbefore
implementingsomeorganizationalorgovernmentalpolicy.Forthemostpart,statisticalinferencemakes
propositionsaboutpopulations,usingdatadrawnfromthepopulationofinterestviasomeformofrandom
sampling.Moregenerally,dataaboutarandomprocessisobtainedfromitsobservedbehaviorduringa
finiteperiodoftime.Givenaparameterorhypothesisaboutwhichonewishestomakeinference,statistical
inferencemostoftenuses:
astatisticalmodeloftherandomprocessthatissupposedtogeneratethedata,whichisknownwhen
randomizationhasbeenused,and
aparticularrealizationoftherandomprocessi.e.,asetofdata.
Regression
Instatistics,regressionanalysisisastatisticalprocessforestimatingtherelationshipsamongvariables.It
includesmanytechniquesformodelingandanalyzingseveralvariables,whenthefocusisonthe
relationshipbetweenadependentvariableandoneormoreindependentvariables.Morespecifically,
regressionanalysishelpsoneunderstandhowthetypicalvalueofthedependentvariable(or'criterion
variable')changeswhenanyoneoftheindependentvariablesisvaried,whiletheotherindependent
variablesareheldfixed.Mostcommonly,regressionanalysisestimatestheconditionalexpectationofthe
dependentvariablegiventheindependentvariablesthatis,theaveragevalueofthedependentvariable
whentheindependentvariablesarefixed.Lesscommonly,thefocusisonaquantile,orotherlocation
parameteroftheconditionaldistributionofthedependentvariablegiventheindependentvariables.Inall
cases,theestimationtargetisafunctionoftheindependentvariablescalledtheregressionfunction.In
regressionanalysis,itisalsoofinteresttocharacterizethevariationofthedependentvariablearoundthe
regressionfunctionwhichcanbedescribedbyaprobabilitydistribution.
Manytechniquesforcarryingoutregressionanalysishavebeendeveloped.Familiarmethodssuchaslinear
regressionandordinaryleastsquaresregressionareparametric,inthattheregressionfunctionisdefinedin
termsofafinitenumberofunknownparametersthatareestimatedfromthedata.Nonparametricregression
referstotechniquesthatallowtheregressionfunctiontolieinaspecifiedsetoffunctions,whichmaybe
infinitedimensional.
Nonparametricstatistics
Nonparametricstatisticsarestatisticsnotbasedonparameterizedfamiliesofprobabilitydistributions.
Theyincludebothdescriptiveandinferentialstatistics.Thetypicalparametersarethemean,variance,etc.
Unlikeparametricstatistics,nonparametricstatisticsmakenoassumptionsabouttheprobability
distributionsofthevariablesbeingassessed.
Nonparametricmethodsarewidelyusedforstudyingpopulationsthattakeonarankedorder(suchas
moviereviewsreceivingonetofourstars).Theuseofnonparametricmethodsmaybenecessarywhendata
havearankingbutnoclearnumericalinterpretation,suchaswhenassessingpreferences.Intermsoflevels
ofmeasurement,nonparametricmethodsresultin"ordinal"data.
Asnonparametricmethodsmakefewerassumptions,theirapplicabilityismuchwiderthanthe
correspondingparametricmethods.Inparticular,theymaybeappliedinsituationswherelessisknown
abouttheapplicationinquestion.Also,duetotherelianceonfewerassumptions,nonparametricmethods
aremorerobust.
Anotherjustificationfortheuseofnonparametricmethodsissimplicity.Incertaincases,evenwhenthe
useofparametricmethodsisjustified,nonparametricmethodsmaybeeasiertouse.Duebothtothis
simplicityandtotheirgreaterrobustness,nonparametricmethodsareseenbysomestatisticiansasleaving
lessroomforimproperuseandmisunderstanding.
Statistics,mathematics,andmathematicalstatistics
Mathematicalstatisticshassubstantialoverlapwiththedisciplineofstatistics.Statisticaltheoristsstudyand
improvestatisticalprocedureswithmathematics,andstatisticalresearchoftenraisesmathematical
questions.Statisticaltheoryreliesonprobabilityanddecisiontheory.
MathematiciansandstatisticianslikeGauss,Laplace,andC.S.Peirceuseddecisiontheorywithprobability
distributionsandlossfunctions(orutilityfunctions).Thedecisiontheoreticapproachtostatisticalinference
wasreinvigoratedbyAbrahamWaldandhissuccessors,[7][8][9][10][11][12][13]andmakesextensiveuseof
scientificcomputing,analysis,andoptimizationforthedesignofexperiments,statisticiansusealgebraand
combinatorics.
Seealso
Asymptotictheory(statistics)
References
1.Lakshmikantham,,ed.byD.Kannan,...V.(2002).Handbookofstochasticanalysisandapplications.NewYork:
M.Dekker.ISBN0824706609.
2.Schervish,MarkJ.(1995).Theoryofstatistics(Corr.2ndprint.ed.).NewYork:Springer.ISBN0387945466.
3.Freedman,D.A.(2005)StatisticalModels:TheoryandPractice,CambridgeUniversityPress.ISBN9780521
671057
4.Hogg,R.V.,A.Craig,andJ.W.McKean."IntrotoMathematicalStatistics."(2005).
5.Larsen,RichardJ.andMarx,MorrisL."AnIntroductiontoMathematicalStatisticsandItsApplications"
(2012).PrenticeHall.
6.Upton,G.,Cook,I.(2008)OxfordDictionaryofStatistics,OUP.ISBN9780199541454
7.Wald,Abraham(1947).Sequentialanalysis.NewYork:JohnWileyandSons.ISBN0471918067."SeeDover
reprint,2004:ISBN0486439127"
8.Wald,Abraham(1950).StatisticalDecisionFunctions.JohnWileyandSons,NewYork.
9.Lehmann,Erich(1997).TestingStatisticalHypotheses(2nded.).ISBN0387949194.
10.Lehmann,ErichCassella,George(1998).TheoryofPointEstimation(2nded.).ISBN0387985026.
11.Bickel,PeterJ.Doksum,KjellA.(2001).MathematicalStatistics:BasicandSelectedTopics1(Second(updated
printing2007)ed.).PearsonPrenticeHall.
12.LeCam,Lucien(1986).AsymptoticMethodsinStatisticalDecisionTheory.SpringerVerlag.ISBN0387
963073.
13.Liese,FriedrichandMiescke,KlausJ.(2008).StatisticalDecisionTheory:Estimation,Testing,andSelection.
Springer.
Additionalreading
Borovkov,A.A.(1999).MathematicalStatistics.CRCPress.ISBN9056990187
VirtualLaboratoriesinProbabilityandStatistics(Univ.ofAla.Huntsville)(http://www.math.uah.ed
u/stat/)
StatiBot(http://www.trigonella.ch/statibot/english/),interactiveonlineexpertsystemonstatistical
tests.
Retrievedfrom"https://en.wikipedia.org/w/index.php?title=Mathematical_statistics&oldid=721010747"
Categories: Statisticaltheory Actuarialscience
Thispagewaslastmodifiedon19May2016,at07:41.
TextisavailableundertheCreativeCommonsAttributionShareAlikeLicenseadditionaltermsmay
apply.Byusingthissite,youagreetotheTermsofUseandPrivacyPolicy.Wikipediaisa
registeredtrademarkoftheWikimediaFoundation,Inc.,anonprofitorganization.
