实验设计与数据处理:Design of Experiments and Analysis of Variance_第1页
实验设计与数据处理:Design of Experiments and Analysis of Variance_第2页
实验设计与数据处理:Design of Experiments and Analysis of Variance_第3页
实验设计与数据处理:Design of Experiments and Analysis of Variance_第4页
实验设计与数据处理:Design of Experiments and Analysis of Variance_第5页
已阅读5页,还剩68页未读 继续免费阅读

付费下载

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

Chapter10DesignofExperimentsandAnalysisofVarianceOne-WayANOVAF-TestTypesof

RegressionModelsExperimentalDesignsOne-WayAnovaCompletelyRandomizedRandomizedBlockTwo-WayAnovaFactorialOne-WayANOVAF-Test1. TeststheEqualityof2orMore(p)PopulationMeans2. VariablesOneNominalScaledIndependentVariable2orMore(p)TreatmentLevelsorClassificationsOneIntervalorRatioScaledDependentVariable3. UsedtoAnalyzeCompletelyRandomizedExperimentalDesignsOne-WayANOVAF-TestAssumptions1. Randomness&IndependenceofErrorsIndependentRandomSamplesareDrawnforeachcondition2. NormalityPopulations(foreachcondition)areNormallyDistributed3. HomogeneityofVariancePopulations(foreachcondition)haveEqualVariancesOne-WayANOVAF-TestHypothesesH0:

1=

2=

3=...=

pAllPopulationMeansareEqualNoTreatmentEffectHa:NotAll

jAreEqualAtLeast1Pop.MeanisDifferentTreatmentEffectNOT

1

2

...

pOne-WayANOVAF-TestHypothesesH0:

1=

2=

3=...=

pAllPopulationMeansareEqualNoTreatmentEffectHa:NotAll

jAreEqualAtLeast1Pop.MeanisDifferentTreatmentEffectNOT

1

2

...

pXf(X)

1

=

2

=

3Xf(X)

1

=

2

3WhyVariances?ObserveonesamplefromeachtreatmentgroupTheirmeansmaybeslightlydifferentHowdifferentisenoughtoconcludepopulationmeansaredifferent?DependsonvariabilitywithineachpopulationHighervarianceinpopulationhighervarianceinmeansStatisticaltestsareconductedbycomparingvariabilitybetweenmeanstovariabilitywithineachsampleTwoPossible

ExperimentOutcomesSametreatmentvariationDifferentrandomvariationACan’trejectequalityofmeans!Rejectequalityofmeans!TwoMorePossible

ExperimentOutcomesSametreatmentvariationDifferentrandomvariationABDifferenttreatmentvariationSamerandomvariationCan’trejectequalityofmeans!RejectReject1. Compares2TypesofVariationtoTestEqualityofMeans2. ComparisonBasisIsRatioofVariances3. IfTreatmentVariationIsSignificantlyGreaterThanRandomVariationthenMeansAreNotEqual4. VariationMeasuresAreObtainedby‘Partitioning’TotalVariationOne-WayANOVA

BasicIdeaOne-WayANOVA

PartitionsTotalVariationOne-WayANOVA

PartitionsTotalVariationTotalvariationOne-WayANOVA

PartitionsTotalVariationVariationduetotreatmentTotalvariationOne-WayANOVA

PartitionsTotalVariationVariationduetotreatmentVariationduetorandomsamplingTotalvariationOne-WayANOVA

PartitionsTotalVariationVariationduetotreatmentVariationduetorandomsamplingTotalvariationSumofSquaresAmongSumofSquaresBetweenSumofSquaresTreatmentAmongGroupsVariationOne-WayANOVA

PartitionsTotalVariationVariationduetotreatmentVariationduetorandomsamplingTotalvariationSumofSquaresWithinSumofSquaresError(SSE)WithinGroupsVariationSumofSquaresAmongSumofSquaresBetweenSumofSquaresTreatment(SST)AmongGroupsVariationTotalVariation

XGroup1Group2Group3Response,XTreatmentVariation

X

X3

X2

X1Group1Group2Group3Response,XRandom(Error)Variation

X2

X1

X3Group1Group2Group3Response,XSS=SSE+SSTButThus,SS=SSE+SSTOne-WayANOVAF-Test

TestStatistic1. TestStatisticF=MST/MSE

MSTIsMeanSquareforTreatmentMSEIsMeanSquareforError2. DegreesofFreedom

1=p-1

2=n-pp=#Populations,Groups,orLevelsn=TotalSampleSizeOne-WayANOVA

SummaryTableSourceofVariationDegreesofFreedomSumofSquaresMeanSquare(Variance)FTreatmentp-1SSTMST=SST/(p-1)MSTMSEErrorn-pSSEMSE=SSE/(n-p)Totaln-1SS(Total)=SST+SSETheFdistributionTwoparametersincreasingeitheronedecreasesF-alpha(exceptforv2<3)I.e.,thedistributiongetssmushedtotheleft

SeeSection9.5

F

v1v2(,)0FOne-WayANOVAF-TestCriticalValue

Ifmeansareequal,F=MST/MSE

1.OnlyrejectlargeF!AlwaysOne-Tail!Fapnp(,)

10RejectH0DoNotRejectH0F©1984-1994T/MakerCo.One-WayANOVAF-TestExampleAsproductionmanager,youwanttoseeif3fillingmachineshavedifferentmeanfillingtimes.Youassign15similarlytrained&experiencedworkers,5permachine,tothemachines.Atthe.05level,isthereadifferenceinmeanfillingtimes?

Mach1 Mach2

Mach3

25.40 23.40 20.00

26.31 21.80 22.20

24.10 23.50 19.75

23.74 22.75 20.60

25.10 21.60 20.40F03.89One-WayANOVAF-Test

SolutionH0:

1=

2=

3Ha:NotAllEqual

=.05

1=2

2=12CriticalValue(s):TestStatistic:Decision:Conclusion:Rejectat

=.05ThereIsEvidencePop.MeansAreDifferent

=.05FMSTMSE

2358209211256...SummaryTable

SolutionFromComputerSourceofVariationDegreesofFreedomSumofSquaresMeanSquare(Variance)FTreatment(Machines)3-1=247.164023.582025.60Error15-3=1211.0532.9211Total15-1=1458.2172Reminder:AssumptionsforEqualityofMeansTestIndependentrandomsamplesfromeachpopulationAllpopulationprobabilitiesarenormallydistributedAllpopulationshaveequalvariances (Teststartswithassumptionofequalmeansaswell,butthatmayberejectedasaresultofthetest)Exercise10.26

|Summaryofvaluecondition|MeanStd.Dev.Freq.------------+------------------------------------1|30.6420.035438502|26.21428623.701946423|15.1276615.70324947------------+------------------------------------Total|24.05755420.878451139Exercise10.26

AnalysisofVarianceSourceSSdfMSFProb>F------------------------------------------------------------------------Betweengroups6109.714123054.857057.690.0007Withingroups54045.8255136397.395776------------------------------------------------------------------------Total60155.5396138435.909707Bartlett'stestforequalvariances:chi2(2)=7.1931Prob>chi2=0.027One-WayANOVAF-Test

ThinkingChallengeYou’reatrainerforMicrosoftCorp.Isthereadifferenceinmeanlearningtimesof12peopleusing4differenttrainingmethods(

=.05)?

M1

M2

M3

M4

10 11 13 18

9 16 8 23

5 9 9 25Usethefollowingtable. ©1984-1994T/MakerCo.SummaryTable

(PartiallyCompleted)SourceofVariationDegreesofFreedomSumofSquaresMeanSquare(Variance)FTreatment(Methods)348Error80TotalF04.07One-WayANOVAF-Test

Solution*H0:

1=

2=

3=

4Ha:NotAllEqual

=.05

1=3

2=8CriticalValue(s):TestStatistic:Decision:Conclusion:Rejectat

=.05ThereIsEvidencePop.MeansAreDifferent

=.05FMSTMSE

11610116.SummaryTable

Solution*SourceofVariationDegreesofFreedomSumofSquaresMeanSquare(Variance)FTreatment(Methods)4-1=334811611.6Error12-4=88010Total12-1=1142810.26:condition1vs.2Two-samplettestwithequalvariances------------------------------------------------------------------------------Group|ObsMeanStd.Err.Std.Dev.[95%Conf.Interval]---------+--------------------------------------------------------------------1|5030.642.83343920.0354424.9459936.334012|4226.214293.6572923.7019518.8282433.60033---------+--------------------------------------------------------------------combined|9228.619572.27025321.775524.1099933.12914---------+--------------------------------------------------------------------diff|4.4257144.559216-4.63196313.48339------------------------------------------------------------------------------Degreesoffreedom:90Ho:mean(1)-mean(2)=diff=0Ha:diff<0Ha:diff!=0Ha:diff>0t=0.9707t=0.9707t=0.9707P<t=0.8329P>|t|=0.3343P>t=0.167110.26condition2vs.3Two-samplettestwithequalvariances------------------------------------------------------------------------------Group|ObsMeanStd.Err.Std.Dev.[95%Conf.Interval]---------+--------------------------------------------------------------------2|4226.214293.6572923.7019518.8282433.600333|4715.127662.29055415.7032510.5170119.73831---------+--------------------------------------------------------------------combined|8920.359552.17653320.5333716.0341524.68495---------+--------------------------------------------------------------------diff|11.086634.2207672.69739419.47586------------------------------------------------------------------------------Degreesoffreedom:87Ho:mean(2)-mean(3)=diff=0Ha:diff<0Ha:diff!=0Ha:diff>0t=2.6267t=2.6267t=2.6267P<t=0.9949P>|t|=0.0102P>t=0.0051MultipleComparisonsProblemP{Atleastoneofpintervalsfailstocontainthetruedifference}=1–P{Allcintervalscontainthetruedifferences}=1–(1-alpha)c>alphaIfcomparingmanypairs,needgreaterconfidenceforanyoneofthemthanyouwouldforrejectingequalityofanyonepairMultipleComparisonsProcedure1. TellsWhichPopulationMeansAreSignificantlyDifferentExample:

1=

2

32. PostHocProcedureDoneAfterRejection

ofEqualMeansin

ANOVAOutputFromManyStatisticalcomputerPrograms–variousversions(Tukey,Bonferroni,etc.)10.26MultipleComparisons

(Bonferroni)RowMean-|ColMean|12---------+----------------------2|-4.42571|0.872|3|-15.5123-11.0866|0.0010.029RandomizedBlockDesignTypesof

RegressionModelsExperimentalDesignsOne-WayAnovaCompletelyRandomizedRandomizedBlockTwo-WayAnovaFactorialRandomizedBlockDesign1. ExperimentalUnits(Subjects)AreAssignedRandomlytoBlocksBlocksareAssumedHomogeneous2. OneFactororIndependentVariableofInterest2orMoreTreatmentLevelsorClassifications3.OneBlockingFactorRandomizedBlockDesignFactorLevels:(Treatments)A,B,C,D

ExperimentalUnits

TreatmentsarerandomlyassignedwithinblocksBlock1ACDBBlock2CDBABlock3BADC

...............BlockbDCABRandomizedBlockF-Test1. TeststheEqualityof2orMore(p)PopulationMeans2. VariablesOneNominalScaledIndependentVariable2orMore(p)TreatmentLevelsorClassificationsOneNominalScaledBlockingVariableOneIntervalorRatioScaledDependentVariable3. UsedwithRandomizedBlockDesignsRandomizedBlockF-TestAssumptions1. NormalityProbabilityDistributionofeachBlock-TreatmentcombinationisNormal2. HomogeneityofVarianceProbabilityDistributionsofallBlock-TreatmentcombinationshaveEqualVariancesRandomizedBlockF-TestHypothesesH0:

1=

2=

3=...=

pAllPopulationMeansareEqualNoTreatmentEffectHa:NotAll

jAreEqualAtLeast1Pop.MeanisDifferentTreatmentEffect

1

2

...

pIsWrong

RandomizedBlockF-TestHypothesesH0:

1=

2=

3=...=

pAllPopulationMeansareEqualNoTreatmentEffectHa:NotAll

jAreEqualAtLeast1Pop.MeanisDifferentTreatmentEffect

1

2

...

pIsWrong

Xf(X)

1

=

2

=

3Xf(X)

1

=

2

3TheFRatioforRandomizedBlockDesignsSS=SSE+SSB+SSTRandomizedBlockF-Test

TestStatistic1. TestStatisticF=MST/MSEMSTIsMeanSquareforTreatmentMSEIsMeanSquareforError2. DegreesofFreedom

1=p-1

2=n–b–p+1p=#Treatments,b=#Blocks,n=TotalSampleSizeRandomizedBlockF-TestCriticalValue

Ifmeansareequal,F=MST/MSE

1.OnlyrejectlargeF!AlwaysOne-Tail!Fapnp(,)

10RejectH0DoNotRejectH0F©1984-1994T/MakerCo.RandomizedBlockF-TestExampleYouwishtodeterminewhichoffourbrandsoftireshasthelongesttreadlife.Yourandomlyassignoneofeachbrand(A,B,C,andD)toatirelocationoneachof5cars.Atthe.05level,isthereadifferenceinmeantreadlife?

TireLocationBlockLeftFrontRightFrontLeftRearRightRearCar1A:42,000C:58,000B:38,000D:44,000Car2B:40,000D:48,000A:39,000C:50,000Car3C:48,000D:39,000B:36,000A:39,000Car4A:41,000B:38,000D:42,000C:43,000Car5D:51,000A:44,000C:52,000B:35,000F03.49RandomizedBlockF-Test

SolutionH0:

1=

2=

3=

4Ha:NotAllEqual

=.05

1=3

2=12CriticalValue(s):TestStatistic:Decision:Conclusion:Rejectat

=.05ThereIsEvidencePop.MeansAreDifferent

=.05F=11.9933Exercise10.47

Whatisthepurposeofblockingonweeksinthisstudy?c.Arethemeannumberofwalkersdifferentamongthepromptingconditions?d.Whichpairwisemeansaresignificantlydifferent?e.Whatassumptionsarerequiredfortheanalysisincandd?FactorialExperimentsTypesof

RegressionModelsExperimentalDesignsOne-WayAnovaCompletelyRandomizedRandomizedBlockTwo-WayAnovaFactorialFactorialDesign1. ExperimentalUnits(Subjects)AreAssignedRandomlytoTreatmentsSubjectsareAssumedHomogeneous2. TwoorMoreFactorsorIndependentVariablesEachHas2orMoreTreatments(Levels)3. AnalyzedbyTwo-WayANOVAFactorialDesign

Example

TreatmentFactor2(TrainingMethod)FactorLevelsLevel1Level2Level3Level119hr.

20hr.

22hr.

Factor

1(High)11hr.

17hr.

31hr.

(Motivation)Level227hr.

25hr.

31hr.

(Low)29hr.

30hr.

49hr.

Advantages

ofFactorialDesigns1. SavesTime&Efforte.g.,CouldUseSeparateCompletelyRandomizedDesignsforEachVariable2. ControlsConfoundingEffectsbyPuttingOtherVariablesintoModel3. CanExploreInteractionBetweenVariablesTwo-WayANOVATypesof

RegressionModelsExperimentalDesignsOne-WayAnovaCompletelyRandomizedRandomizedBlockTwo-WayAnovaFactorialTwo-WayANOVA1. TeststheEqualityof2orMorePopulationMeansWhenSeveralIndependentVariablesAreUsed2. SameResultsasSeparateOne-WayANOVAonEachVariableButInteractionCanBeTested3. UsedtoAnalyzeFactorialDesignsTwo-WayANOVAAssumptions1. NormalityPopulationsareNormallyDistributed2. HomogeneityofVariancePopulationshaveEqualVariances3. IndependenceofErrorsIndependentRandomSamplesareDrawnTwo-WayANOVA

DataTableXijk

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论