You are on page 1of 47

EffectsofBreakfastSizeonStressLevels

AnalyticalReport
KyleSheth
D/CStatistics
Period6B
5/23/16
Abstract:
Thisreportexaminesapotentialassociationbetweenthestressrightbeforeastressful
eventandthesizeofbreakfasteatenthatmorning.ThepopulationinthisexperimentisallLASA
students,andasampleof31studentswasrandomlyacquiredinordertoconductanANOVAtest
tocomparethemeanstresslevelsofeachbreakfastsizegroup.Inthisexperiment,the
participantswereaskedtoeatacertainsizebreakfastthatwasintermsof,butnotlimitedto,
eggs(small:02eggs,medium:23eggs,large:4+eggs)onthemorningofastressfulevent,and
foramajorityofthesestudents,thestressfuleventwasanAdvancedPlacementExam.These31
studentswererandomlyassignedintothe3groupsasfollows:11eatingalargebreakfast,12
eatingamediumbreakfast,and8eatingasmallbreakfast.Rightbeforetakingthetest,theywere
askedtoratetheirstressonascalefrom1(low)to5(high).ResultsarecalculatedusingRStudio
andapvalueiscomparedtothesignificancelevelof0.05inordertodeterminethestateofthe
nullhypothesis,whichsaysthatthetruemeanstresslevelsarethesameacrossall3breakfast
sizegroups.


Sheth1

Table of Contents
TableofContents

Introduction

SampleCollection

ExperimentalProcedure

DataAnalysis

Conclusion

WorksCited

AppendixAAttemptsatafullandrandomizedsample

AppendixBData

AppendixCParticipantSignatures

10

AppendixDProofofRandomization

12

AppendixERStudioOutput

30

AppendixFFigures

32

AppendixGWebsiteLink

33

AppendixHRevisedInquiryPitch

34

AppendixIOriginalInquiryPitch

37

AppendixJReportRoughDraft

40

AppendixKPeerReviewFeedbackForm

45


Sheth2

Introduction
Everysingleday,manypeopleareparticipatinginsomekindofimportanteventthat
requiresfocus,rangingfromjobinterviewstoexams.Often,peoplewonderhowtoachievean
optimalperformanceonsuchevents,andformost,eatingalargebreakfastseemstobea
commonsolution,asitgivesonealargerstoreofenergythroughoutthedayaswellasfurther
pushesoffthedistractionofanemptystomach(Chaplin).Therearemanypeoplewhobelieve
thatnomatterwhatlengthstheygotolessenstressfromtheevent,itpersists.Eatingalarger
breakfastmayhavebeenshowntoalleviatetheintensityofbodilydistractions(e.g.hunger)and
makeiteasiertofocuson(forexample)atestathandbutcanithelpalleviatethestressthat
preventsamindfromfullyfocusing?Inthisexperiment,apossibleassociationbetween
breakfastsizeandstresslevelforaneventwillbeinvestigatedinstudentsattheLiberalArtsand
ScienceAcademy(LASA).

Sample Collection
TheLASAstudentbodyiscomposedsimplyofstudentswhoarecurrentlyregisteredat
LASA.MysamplingframewasarosterofallregisteredLASAstudents,andtoselectthe
sample,therosterwasplacedintoalistrandomizerandthesamplewasthefirst33studentsin
thelistusingRandom.orgslistrandomizer(True).Thefirst11studentsofthissamplewere
assignedthelargebreakfastsize,thesecond11wereassignedthemediumbreakfastsize,andthe
last11wereassignedthesmallbreakfastsize.However,duetoanimmenseamountof
nonresponsebias(AppendixA),thetotalsampleendedupwithonly31students,andthe
numberofparticipantsineachbreakfastsizegroupwasnotthesameaswasproposed.


Sheth3

ThissamplemaynothavebeenrepresentativeoftheLASAstudentbodyasawholedue
tothevoluntaryresponsebiasaswellasthenonresponsebias(somestudentsrejectedto
participateintheexperiment,andsomestudentsdidntreplyaltogether).Areasonthatthis
wouldhaveanimpactonthesampleisthatthosewhovolunteeredtoparticipatecouldhave
somelurkingfactorthatmayaffectthewaythatbreakfastsizeaffectstheirstresslevels,and
thosewhodidntreplyordecidednottoparticipatemadeitsothatcertaingroupsofindividuals
hadlessofarepresentationinthesample.

Experimental Procedure
Eachparticipantselfdeterminedwhattheirstressfuleventwouldbe,butmostofthese
endedupbeingAdvancedPlacementExams,becausetheyoccurredconcurrentlywiththedata
collection.Onthedayofaparticipantsevent,theywouldeatthebreakfastsizethatwas
randomlyassignedtothem(AppendixA).Afewminutesbeforetheireventbegan,theirstress
levelwouldbeselfgaugedonascalefrom1(lowstress)to5(highstress),theresultsofwhich
areshowninAppendixB.

Data Analysis
AnANOVAtestwasconductedonthedatabecauseoneofthevariablesbeingtestedis
anintervalandquantitativeone(thestresslevel,onascalefrom1(low)5(high)),andtheotheris
anominalandcategoricalvariable(breakfastsize:small,medium,andlarge).Theconditionsfor
anANOVAtestarethatthesamplewasrandomlychosenfromthepopulation,thedatapoints
areindependentofeachother,thevariancesinthepopulationdistributionsforthestresslevelsof


Sheth4

eachofthethreegroupswereapproximatelyequal,andfinally,ifthepopulationdistributionsor
samplingdistributionsofthestresslevelsforeachgroupwereapproximatelynormal.The
samplewasrandomlyselectedfromthepopulationandrandomlydividedintogroupsforthistest
topertaintoanexperimentinsteadofanobservationalstudy,sotherandomnessconditionis
met.Theindependenceconditionisassumedtobetrue,asitisassumedthatthereareatleast10x
thesamplesizeofthatcertainindividualbeingtestedinthepopulationitself.Intermsofthis
test,thatmeansitisassumedthatthereareatleast110studentswhoeatlargebreakfastsbeforea
stressfulevent,atleast120studentswhoeatamediumbreakfastbeforeastressfulevent,andat
least80studentswhoeatasmallbreakfastornobreakfastbeforeastressfuleventatLASA.
Astandardwayofcheckingifthepopulationvariancesareapproximatelyequalistotake
thelargestvarianceanditshouldfallinbetweenthesmallestvarianceanddoublethesmallest
variance.However,theequalvariancesconditionisnotmetwiththisdata,becausethevariance
instresslevelforthelargebreakfastsizewasapproximately0.273(whendoubled,itis0.546)
andthevarianceinstresslevelforthesmallbreakfastwas0.857,whichwasnotinsideofthe
bounds:0.27273<0.857

0.546(AppendixE).Thenormalityconditionwasalsonotmet,asat
leastoneofthesamplesdistributionsofstresslevelwasskewed,asshowninthesomeofthe
nonsymmetricboxplotsandhistograms(AppendixF)anditssamplesizewasnotgreaterthan
30.Becausetheseassumptionswerenotmet,theconclusionsofthisstatisticaltestmaybe
subjecttoquestionastheresultsmayhavebeenbiasedbytheseviolations.
ThenullhypothesisofthisANOVAtestisthatthetruepopulationmeanstresslevelsare
thesameandthealternativehypothesisisthatleastoneofthemeanstresslevelsisdifferentfrom
theothers.Theselectedalphasignificancelevelforthistestwas0.05(probabilityofafalse


Sheth5

positiveTypeIError)becauseitisacommonstandardtofollowwhenconductingstatistical
proceduresinageneralfield(asopposedtocertainfieldsofstudywherealowerchanceofa
falsepositiveisneeded).ConductingthetestinRStudiowithadegreesfreedombetweengroups
of2andadegreesfreedomwithingroupsof28,thetestresultedinasmallFstatisticofonly
0.876andalargepvalueof0.428.

Conclusion
Becausetheresultingpvalueofthetestwasgreaterthanthesignificancelevelof0.05,
thenullhypothesis,whichsaysthatthetruepopulationmeanstresslevelsarethesame,failsto
berejected.Therefore,thereisnotenoughevidencetosupporttheclaimthatstresslevelbefore
aneventisassociatedwiththesizeofthebreakfasteatenthemorningoftheevent.Furthermore,
aTukeyHSDposthoctestwasconductedtoconfirmthattherewerenosignificantdifferences
betweenanyoftheindividualgroups,andalloftheadjustedpvalueswerestillgreaterthanour
initialsignificantlevel(AppendixE).Therefore,nosignificantdifferenceswerefoundthere,so
thenullhypothesisisstillnotrejected.Becauseofthis,itispossibleforthisexperimenttohave
resultedaTypeIIError(falsenegative),whichisinthecaseofthisexperiment,theresultsfrom
thedatacouldincorrectlyconcludethatbreakfastsizehasnoassociationwithstresslevelright
beforeaneventwheninrealityitdoesduetorandomsamplingerror.
Amajortakeawayfromthisexperimentisthatmeetingalloftheconditionsforan
ANOVAtestrequiresalargeroverallsamplesizesothateachindividualgroupbeingtested
containsmorethan30individualsforthenormalitycondition(accordingtotheCentralLimit
Theorem,thesamplingdistributionofthesamplemeanswithasamplesizeofatleast30willbe


Sheth6

approximatelynormal,regardlessofthepopulationdistributionorsampledistribution'sshape),
andalsosothatthevariancesofeachgroupssamplingdistributionisclosertotheothers
becausetheCLTsaysthatthehigherthesamplesizeis,themoreclusteredaroundthemeanand
narrowerthehistogramofthesamplingdistributionis.Therefore,ifthisexperimentwastobe
redone,thegoalsamplesizeforeachgroupwouldbemorethan30,butthetotalsamplesizefor
allofthegroupscombinedshouldbelessthan10%ofthetotalpopulation.
Anotherimprovementtothisexperimentcouldbe(accompanyingalargersamplesize)a
largerscaleforthequantitativevariable(forexample,010stresslevelinsteadof15),asthe
histogramandboxplotsforthecollecteddata(AppendixF)donotshowmuchvariation,withall
ofthequantitativedatahaving3options,essentiallyactingasnotonlyaquantitativevariable,
butalsoacrudecategoricalvariable.Becauseofthis,insteadofaskingparticipantstorate
themselvesonabroaderscale,achisquaredtestofindependencecouldbeconducted,withone
categoricalvariablebeingthebreakfastsize(small,medium,large)andtheotherbeinga
categoricalstresslevel(nostress,littlestress,moderatestress,oralotofstress,verysimilarto
the15scale).Thiscouldbebetterbecauseavisualrepresentationofthedatacouldbeabarplot
insteadofahistogramandnormalityofthedata(difficulttoachievewithasubjective
quantitativemeasurementlikestress)isnolongerarequiredconditionforthehypothesistest,
andisinsteadreplacedbyexpectedcountsconditionsthatcaneasilybemetbyincreasingthe
samplesize(asmentionedabove).Regardlessofthesefutureimprovements,theexperimentin
itscurrentstateaswellasthecollecteddatatogetherdidnotprovideenoughevidencetosupport
thateatingalargerbreakfastcanreducetheamountofstressbeforeanevent.


Sheth7

Works Cited
Chaplin,Katherine,andAndrewP.Smith.BreakfastandSnacks:AssociationswithCognitive
Failures,MinorInjuries,AccidentsandStress.
Nutrients
3.5(2011):515528.
PMC
.
Web.16Apr.2016.
"TrueRandomNumberService."
RANDOM.ORG
.RANDOM.ORG,n.d.Web.15Apr.2016.


Sheth8

Appendix A -
Attempts at a full and randomized sample:
Duetoalotofnonresponsebias(andvoluntaryresponsebias,participationinthe
experimentwasalsomadevoluntaryincaseanindividualfeltthatparticipationwouldaffect
theirperformanceintheeventnegatively),2listsofalternateswerecreatedtoreplace
individualsinthesamplewhoneverresponded/optedout,madeof2separaterandomly
generatedsequencesof33additionalindividuals,selectedfromthosewhowerenotpartofthe
first33individualsintheoriginalrandomizedlist(AppendixD).Breakfastsizesinthefirstlist
ofalternateswasdeterminedbygoingdownthelistandeverythirdpersonstartingfromthefirst
individualwasalarge,everythirdpersonstartingfromthesecondonewasamedium,andevery
thirdpersonstartingfromthethirdonewasasmall.Inthelastlist,breakfastsizewasrandomly
selectedforeachindividual.However,outofthese99peoplecontactedfortheexperiment
(about10%ofmypopulationof1010students),only31individualsagreedtoparticipateinthe
experiment,withthefollowingdistributionofbreakfastsizes:11large,12medium,and8small.
Fortheinitialpopulation,becausethelistorderitselfwasrandomized,theassignmentofthe
larges,mediums,andsmallswastechnicallyrandombecauseitwasdeterminedbytherandom
orderingofthelist.Thebreakfastsizeassignmentsforthesecondlistfollowsthesame
reasoning.Thefinallistssizeassignmentswereinherentlyrandom.


Sheth9

Appendix B - Data
Breakfast Stress
Large

Large

Large

Large

Large

Large

Large

Large

Large

Large

Large

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Medium

Small

Small

Small

Small

Small

Small

Small

Small


Sheth10

Appendix C - Participant Signatures


SierraObermoellerGilmer

NilanBhakta

JacksonRoberts

MarcMartinez

ElysaNaranjo

NickHammel

CiaraMcDaniel

NickBlacklock

(doneviaFacebookDoodle)
DerianGolden

ZacharyKosub


Sheth11

ChristineGarner

RileyPruden

MaxHenneke

UlanNotzon

EzraSerrins

AndiFeddeler

DanielPyrek

AnastasiaBarnett


Sheth12

Appendix D - Randomization Evidence


Randomizedlistusedforinitialsampleof33andlistofindexesusedforalternates:


Sheth13


Sheth14


Sheth15


Sheth16


Sheth17


Sheth18


Sheth19


Sheth20


Sheth21


Sheth22


Sheth23


Sheth24


Sheth25


Sheth26


Sheth27


Sheth28

Firstrandomizedlistofalternatesindexes:


Sheth29

Secondrandomizedlistofalternatesindexes:


Sheth30

Appendix E - RStudio Output


>#Importandsubsetdata
>data<read.csv("~/FinalProjectData(Compiled)Exportable.csv")
>large<data[data$Breakfast=='Large',]
>medium<data[data$Breakfast=='Medium',]
>small<data[data$Breakfast=='Small',]
>
>#Showhowmanyparticipantsareineachgroup
>table(data$Breakfast)
LargeMediumSmall
11128
>
>#AverageStresslevelforeachgroup
>aggregate(Stress~Breakfast,data,mean)
BreakfastStress
1Large1.545455
2Medium1.833333
3Small2.000000
>
>#VarianceinStressforeachgroup
>aggregate(Stress~Breakfast,data,var)
BreakfastStress
1Large0.2727273
2Medium0.6969697
3Small0.8571429
>
>#Visualizethedata(boxplotsandhistograms)
>hist(small$Stress,main='DistributionofStressLevelsbyBreakfast
Size',xlab='StressLevel',col='#4da34f66',xlim=c(1,5),ylim=c(0,6),breaks=3)
>hist(medium$Stress,col='#ff000066',add=T,breaks=3)
>hist(large$Stress,col='#3c00ff66',add=T,breaks=2)
>
legend(4,5,c('Large','Medium','Small'),pch=c(20,20,20),col=c('#3c00ff66','#ff0000
66','#4da34f66'),cex=2)
>boxplot(data$Stress~data$Breakfast,main="Participant'sStressLevelsby
BreakfastSize",xlab="BreakfastSize",ylab="Stress
Level",col=c('#3c00ff66','#ff000066','#4da34f66'))
>
>#RunANOVAtest
>modelbud<aov(data$Stress~data$Breakfast)
>summary(modelbud)
DfSumSqMeanSqFvaluePr(>F)
data$Breakfast21.0250.51270.8760.428
Residuals2816.3940.5855

>
>#RunaTukeyHSDwithadjustedpvaluesasaposthoctest


Sheth31

>TukeyHSD(modelbud)
Tukeymultiplecomparisonsofmeans
95%familywiseconfidencelevel
Fit:aov(formula=data$Stress~data$Breakfast)
$`data$Breakfast`
difflwruprpadj
MediumLarge0.28787880.50243721.0781950.6439789
SmallLarge0.45454550.42520461.3342950.4187838
SmallMedium0.16666670.69751131.0308450.8825594


Sheth32

Appendix F - Figures


Sheth33

Appendix G - Website Link


http://breakfastandstress.weebly.com


Sheth34

Appendix H - Revised Inquiry Pitch


Itiswellknownthatthemorningbeforeanimportanteventthatrequiresfocus,likea
test,competition,orjobinterview,therearemanyactionsonecancompleteinordertomaintain
optimalperformance.Thisincludeseatingalargebreakfastinordertohavealargerstoreof
energyduringtheeventandtoavoidthedistractionofanemptystomach(Chaplin).However,
somefeelthatthestressaccompanyingsuchaneventisinevitable,regardlessofhowprepared
theyareforit.Inthisexperiment,theinevitabilityofstresswillbeputtothetestagainstvarious
breakfastsizesinordertodetermineifsuchstresscanbereducedbysimplyeatingmorefood.
Soisstresslevelbeforeanimportanteventassociatedwithbreakfastsize?
Thetwovariablesunderinvestigationwillbebreakfastsize,whichwillbeconvertedinto
acategorical/nominalvariable,andstresslevelbeforeatest,whichwillbequantitativeand
measuredasaquantitativeintervalvariable.Breakfastsizewillhavethefollowingcategories:
small,medium,andlarge.Eachcategorywillbeassignedacertainrangeofweightsforthe
breakfast,ascalorieswouldbehardertomeasure.Anexpectedrangeofvaluesforall3
categorieswouldbe1ozto8oz.Stresslevelwillbeonascalefrom1to5,with1beingnostress
and5beingextremelyhighamountsofstress.Expectedvaluesforstressincludesthewhole
scale,butthedistributionwillprobablyfollowmoreofapositivelyskeweddistribution,withless
participantsfeelingmorestressed.
Firstly,theplannedpopulationforthisexperimentisLASAstudentsbecausehavinga
peerconductedexperimentwouldresultinthemostdataforathoroughanalysisthatdoesnot
violatetheassumptionsofthevarioushypothesisteststhatcanbeused.Abroaderpopulationis
notusedduetoavailabilityofdata,randomizationtechniques,andthelogisticsofconvincing


Sheth35

peopleingeneraltochangetheirmorningmealsizeforacompletestranger.Toobtainarandom
sampleofLASAstudents,everystudentwillbeindexedanumberfrom1to~900(totalnumber
ofstudentsatLASA)onaroster.Fromtheseindices,33randomnumberswillbechosen,
regardlessoffactorssuchasclasslevel,(usingthenearlytruerandomnessofrandom.orgs
randomintegergenerator),eachcorrespondingtoarandomstudentchosentoparticipateinthis
experiment.Thoughitishighlylikelythattherosterwillbealphabetizedinadvance,thiswill
notreducetherandomnessofthesamplebecausethevalueofanindexdoesnotaffecthow
likelyitistobepicked,soallofthestudentstheoreticallystillhavethesamechanceofbeing
picked.
These33subjectswillberandomlydividedinto3groupsof11(usingrandom.orgagain),
eachgroupreceivingadifferentlevelofbreakfastsize.Thesebreakfastsizeswillbesomewhat
closetothefollowing:small(12oz),medium(35oz),andlarge(68oz).Thisexperimentwillbe
donejustbeforeAPweek(weekofMay2nd),agenerallystressfuleventformoststudents
(assumingmostofthemaretakingAPs).Eachparticipantwilleatthespecifiedbreakfastsizeon
themorningofthedayoftheirnextAPanduponarrivingtoschool,willbeaskedtoratetheir
currentstresslevel.
Usingthisdata,anANOVAtestwillbeconductedinordertocomparethegroupmeans
(meanstresslevelforeachbreakfastsize).ANOVAisusedinsteadofanindependentsample
ttestbecausetherearemorethan2factorlevels(thethreesizesofbreakfast).Thenull
hypothesisofthetestisthatthetruepopulationmeanstresslevelsarethesameandthe
alternativehypothesisisthatleastoneofthemeanstresslevelsisdifferentfromtheothers.
AnotherreasontheANOVAtestwaschosen,wasthatincaseofarejectionofthenull


Sheth36

hypothesis,aTukeyHSDposthocanalysiscouldbeconductedinRStudioinordertoidentify
whichofthethreebreakfastsizeshasadifferentstresslevelthantheothersandwhetherthereis
agreaterorlesserstresslevelforthesignificantlydifferentmeans.
Overall,thisexperimentwillmeasurethestresslevelofLASAstudentsrightbeforetheir
firstAPthisyear,dividedupbythesizeofthemealtheyate,whichwasrandomlydetermined.
TheresultswillshowwhetherornotadifferentbreakfastsizeaffectsstresslevelsofaLASA
studentbeforeastressfulevent.Someexpectedissuesintheexperimentincludeactually
measuringtheweightoffood,thesubjectivestresslevelsofindividuals,theactualamountof
LASAstudentswhowillactuallytakeAPs,andfinallythesubjectivedifficultiesofeachAP
(whichisrelatedtothesubjectivityofstresslevels).Thefirstissuecanbeaddressedbyweighing
foodonaplate,thensubtractingtheweightoftheplate,whichactuallyisntmuchofaproblem.
However,thefactthatstresslevelsaresubjectiveforindividualsmeansthatthescaleofstress
maybeslightlyinaccurateindepictingatruepsychologicalstresslevel,whichwouldbe
extremelydifficulttodo.Additionally,thisexperimentdrawssamplesfromallLASAstudents,
includingthosenottakingAPs,sosomemayfeelnostressfromAPweekbecausetheyhaveno
teststostressover.TheproblemliesinthefactthatIwouldliketokeepmypopulationasthe
entiretyofLASAstudentsinorderfortheresultsofthisexperimenttoapplytostressfulevents
beyondAPs.Astudentwhoistaking7APsvs.apersonwhoistakingonly1or2APsshould
haveaboutthesamestresslevelforasingleday,becausethestressrelatedquestionwillbe
phrasedsothatthestressperdayismeasuredandnotthestressforallofAPweekandthe
numberofAPsonecantakeinadayonlyrangesfrom02(inclusively).ApersontakingnoAPs


Sheth37

likelyhassomekindoftestduringtheAPweeksthatissimilartoanAP,ifnot,theirstresslevel
willbetakenasis.
Appendix I - Original Inquiry Pitch
Itiswellknownthatthemorningbeforeanimportanteventthatrequiresfocus,likea
test,competition,orjobinterview,therearemanyactionsonecancompleteinordertomaintain
optimalperformance.Thisincludeseatingalargebreakfastinordertohavealargerstoreof
energyduringtheeventandtoavoidthedistractionofanemptystomach(Chaplin).However,
somefeelthatthestressaccompanyingsuchaneventisinevitable,regardlessofhowprepared
theyareforit.Inthisexperiment,theinevitabilityofstresswillbeputtothetestagainstvarious
breakfastsizesinordertodetermineifsuchstresscanbereducedbysimplyeatingmorefood.
Thetwovariablesunderinvestigationwillbebreakfastsize,whichwillbeconvertedinto
acategorical/nominalvariable,andstresslevelbeforeatest,whichwillbequantitativeand
measuredasaquantitativeintervalvariable.Breakfastsizewillbemeasuredinounces(weight)
andnotcaloriesduetodifficultyinprecisemeasurements.Anexpectedrangeofvaluesforthis
wouldbe1ozto8oz.Stresslevelwillbeonascalefrom1to10,with1beingnostressand10
beingextremelyhighamountsofstress.Expectedvaluesforstressincludesthewholescale,but
thedistributionwillprobablyfollowmoreofapositivelyskeweddistribution,withless
participantsfeelingmorestressed.
Firstly,theplannedpopulationforthisexperimentisLASAstudentsbecause,havinga
peerconductedexperimentwouldresultinthemostdataforathoroughanalysisthatdoesnot
violatetheassumptionsofthevarioushypothesisteststhatcanbeused.Abroaderpopulationis
notusedduetoavailabilityofdata,randomizationtechniques,andthelogisticsofconvincing


Sheth38

peopleingeneraltochangetheirmorningmealsizeforacompletestranger.Toobtainarandom
sampleofLASAstudents,everystudentwillbeindexedanumberfrom1to~900(totalnumber
ofstudentsatLASA)onaroster.Fromtheseindices,33randomnumberswillbechosen,
regardlessoffactorssuchasclasslevel,(usingthenearlytruerandomnessofrandom.orgs
randomintegergenerator),eachcorrespondingtoarandomstudentchosentoparticipateinthis
experiment.Thoughitishighlylikelythattherosterwillbealphabetizedinadvance,thiswill
notreducetherandomnessofthesamplebecausethevalueofanindexdoesnotaffecthow
likelyitistobepicked,soallofthestudentstheoreticallystillhavethesamechanceofbeing
picked.
These33subjectswillberandomlydividedinto3groupsof11(usingrandom.orgagain),
eachgroupreceivingadifferentlevelofbreakfastsize.Thesebreakfastsizeswillbesomewhat
closetothefollowing:small(12oz),medium(35oz),andlarge(68oz).Thisexperimentwillbe
donejustbeforeAPweek(weekofMay2nd),agenerallystressfuleventformoststudents
(assumingmostofthemaretakingAPs).Eachparticipantwilleatthespecifiedbreakfastsizeon
themorningofthedayoftheirfirstAPanduponarrivingtoschool,willbeaskedtoratetheir
currentstresslevel.
Usingthisdata,anANOVAtestwillbeconductedinordertocomparethegroupmeans
(meanstresslevelforeachbreakfastsize).ANOVAisusedinsteadofanindependentsample
ttestbecausetherearemorethan2factorlevels(thethreesizesofbreakfast).Thenull
hypothesisofthetestisthatthetruepopulationmeanstresslevelsarethesameandthe
alternativehypothesisisthatleastoneofthemeanstresslevelsisdifferentfromtheothers.
AnotherreasontheANOVAtestwaschosen,wasthatincaseofarejectionofthenull


Sheth39

hypothesis,aTukeyHSDposthocanalysiscouldbeconductedinRStudioinordertoidentify
whichofthethreebreakfastsizeshasadifferentstresslevelthantheothersandwhetherthereis
agreaterorlesserstresslevelforthesignificantlydifferentmeans.
Overall,thisexperimentwillmeasurethestresslevelofLASAstudentsrightbeforetheir
firstAPthisyear,dividedupbythesizeofthemealtheyate,whichwasrandomlydetermined.
TheresultswillshowwhetherornotadifferentbreakfastsizeaffectsstresslevelsofaLASA
studentbeforeastressfulevent.Someexpectedissuesintheexperimentincludeactually
measuringtheweightoffood,thesubjectivestresslevelsofindividuals,theactualamountof
LASAstudentswhowillactuallytakeAPs,andfinallythesubjectivedifficultiesofeachAP
(whichisrelatedtothesubjectivityofstresslevels).Thefirstissuecanbeaddressedbyweighing
foodonaplate,thensubtractingtheweightoftheplate,whichactuallyisntmuchofaproblem.
However,thefactthatstresslevelsaresubjectiveforindividualsmeansthatthescaleofstress
maybeslightlyinaccurateindepictingatruepsychologicalstresslevel,whichwouldbe
extremelydifficulttodo.Additionally,thisexperimentdrawssamplesfromallLASAstudents,
includingthosenottakingAPs,sosomemayfeelnostressfromAPweekbecausetheyhaveno
teststostressover.TheproblemliesinthefactthatIwouldliketokeepmypopulationasthe
entiretyofLASAstudentsinorderfortheresultsofthisexperimenttoapplytostressfulevents
beyondAPs.


Sheth40

Appendix J - Report Rough Draft


Introduction
Everysingleday,manypeopleareparticipatinginsomekindofimportanteventthat
requiresfocus,rangingfromjobinterviewstoexams.Often,peoplewonderhowtoachievean
optimalperformanceonsuchevents,andformost,eatingalargebreakfastseemstobea
commonsolution,asitgivesonealargerstoreofenergythroughoutthedayaswellasfurther
pushingoffthedistractionofanemptystomach(Chaplin).Therearemanypeoplewhobelieve
thatnomatterwhatlengthstheygotolessenstressfromtheevent,itpersists.Eatingalarger
breakfastmayhavebeenshowntoalleviatetheintensityofbodilydistractions(e.g.hunger)and
makeiteasiertofocuson(forexample)atestathandbutcanithelpalleviatethestressthat
preventsamindfromfullyfocusing?Inthisexperiment,apossibleassociationbetween
breakfastsizeandstresslevelforaneventwillbeinvestigatedinLASAstudents.

Sample Collection
TheLASAstudentbodyiscomposedsimplyofstudentswhoareregisteredatLASA.My
samplingframewasarosterofallregisteredLASAstudents,andtoselectthesample,theroster
wasplacedintoalistrandomizerandthesamplewasthefirst33studentsinthelist(True).The
first11studentsofthissamplewereassignedthelargebreakfastsize,thesecond11were
assignedthemediumbreakfastsize,andthelast11wereassignedthesmallbreakfastsize.
However,duetoanimmenseamountofnonresponsebias(AppendixA),thetotalsampleended
upwithonly31students,andthenumberofparticipantsineachbreakfastsizegroupwasnotthe
same,aswasproposed.


Sheth41

ThissamplemaynothavebeenrepresentativeoftheLASAstudentbodyasawholedue
tothevoluntaryresponsebiasaswellasthenonresponsebias(somestudentsrejectedto
participateintheexperiment,andsomestudentsdidntreplyaltogether).Areasonthatthis
wouldhaveanimpactonthesampleisthatthosewhovolunteeredtoparticipatecouldhave
somelurkingfactorthatmayaffectthewaythatbreakfastsizeaffectstheirstresslevels,and
thosewhodidntreplyordecidednottoparticipatemadeitsothatcertaingroupsofindividuals
hadlessofarepresentationinthesample.

Data Analysis
AnANOVAtestwasconductedonthedata(AppendixB)becausetheofthetwo
variablesforwhichanassociationisbeingtestedfor,oneisanintervalandquantitativevariable
(thestresslevel,onascalefrom1(low)5(high)),andtheotheroneisanominalandcategorical
variable(breakfastsize).TheconditionsforanANOVAtestarethatthesamplewasrandomly
chosenfromthepopulation,thedatapointsareindependentofeachother,thevariancesinthe
populationdistributionsforthestresslevelsofeachofthethreegroupswereapproximately
equal,andfinally,ifthepopulationdistributionsorsamplingdistributionsofthestresslevelsfor
eachgroupwereapproximatelynormal.Thesamplewasrandomlyselectedfromthepopulation
andrandomlydividedintogroupsforthistesttopertaintoanexperimentinsteadofan
observationalstudy,sotherandomnessconditionismet.Theindependenceconditionisassumed
tobetrue,asitisassumedthatthereareatleast10xthesamplesizeofthatcertainindividual
beingtestedinthepopulationitself.Intermsofthistest,thatmeansitisassumedthatthereareat
least110studentswhoeatlargebreakfastsbeforeastressfulevent,atleast120studentswhoeat


Sheth42

amediumbreakfastbeforeastressfulevent,andatleast80studentswhoeatasmallbreakfastor
nobreakfastbeforeastressfuleventatLASA.
Inorderfortheequalvariancesconditiontobemet,astandardwayofcheckingifthe
populationvariancesareapproximatelyequalistotakethelargestvarianceanditshouldfallinto
therangeofthesmallestvarianceanddoublethesmallestvariance.However,theequal
variancesconditionisnotmetwiththisdata,becausethevarianceinstresslevelforthelarge
breakfastsizewasapproximately0.27273(whendoubled,itis0.5455)andthevarianceinstress
levelforthesmallbreakfastwas0.857,whichwasnotinsideofthebounds:0.27273<0.857

0.5455(AppendixD).Thenormalityconditionwasalsonotmet,asatleastoneofthesamples
distributionsofstresslevelwasskewed,asshowninthesomeofthenonsymmetricboxplots
andhistograms(AppendixE)anditssamplesizewasnotgreaterthan30.Becausethese
assumptionswerenotmet,theconclusionsofthisstatisticaltestmaybesubjecttoquestionas
theresultsmayhavebeenbiasedbytheseviolations.
ThenullhypothesisofthisANOVAtestisthatthetruepopulationmeanstresslevelsare
thesameandthealternativehypothesisisthatleastoneofthemeanstresslevelsisdifferentfrom
theothers.Theselectedalphasignificancelevelforthistestwasselectedtobe0.05(5%chance
ofTypeIError)becauseitisacommonstandardtofollowwhenconductingstatistical
procedures.ActuallyconductingthetestinRStudiowithadegreesfreedombetweengroupsof2
andadegreesfreedomwithingroupsof28,thetestresultedinasmallFstatisticofonly0.876
andapvalueof0.428.


Sheth43

Conclusion
Becausetheresultingpvalueofthetestwasgreaterthanthesignificancelevelof0.05,
thenullhypothesis,whichsaysthatthetruepopulationmeanstresslevelsarethesame,failsto
berejected.Therefore,thereisnotenoughevidencetosupporttheclaimthatstresslevelbefore
aneventisassociatedwiththesizeofthebreakfasteatenthemorningoftheevent.Furthermore,
aTukeyHSDposthoctestwasconductedtoconfirmthattherewerenosignificantdifferences
betweenanyoftheindividualgroups,andalloftheadjustedpvalueswerestillgreaterthanour
initialsignificantlevel(AppendixD),sonosignificantdifferenceswerefoundthere,sothenull
hypothesisisstillnotrejected.Becauseofthis,itispossibleforthisexperimenttohaveresulted
aTypeIIError(falsenegative),whichisinthecaseofthisexperiment,theresultsfromthedata
couldincorrectlyconcludethatbreakfastsizehasnoassociationwithstresslevelrightbeforean
eventwheninrealityitdoesduetorandomsamplingerror.
Amajortakeawayfromthisexperimentisthatmeetingalloftheconditionsforan
ANOVAtestrequiresalargeroverallsamplesizesothateachindividualgroupbeingtested
containsmorethan30individualsforthenormalitycondition(accordingtotheCentralLimit
Theorem,thesamplingdistributionofthesamplemeanswithasamplesizeofatleast30willbe
approximatelynormal,regardlessofthepopulationdistributionorsampledistribution'sshape),
andalsosothatthevariancesofeachgroupssamplingdistributionisclosertotheothers
becausetheCLTsaysthatthehigherthesamplesizeis,themoreclusteredaroundthemeanand
narrowerthehistogramofthesamplingdistributionis.Therefore,ifthisexperimentwastobe
redone,thegoalsamplesizeforeachgroupwouldbemorethan30,butthetotalsamplesizefor
allofthegroupscombinedshouldbelessthan10%ofthetotalpopulation.


Sheth44

Anotherimprovementtothisexperimentcouldbe(accompanyingalargersamplesize)a
largerscaleforthequantitativevariable(forexample,010stresslevelinsteadof15),asthe
histogramandboxplotsforthecollecteddata(AppendixE)donotshowmuchvariation,withall
ofthequantitativedatahaving3options,essentiallyactingasnotonlyaquantitativevariable,
butalsoacrudecategoricalvariable.Becauseofthis,insteadofaskingparticipantstorate
themselvesonabroaderscale,achisquaredtestofindependencecouldbeconducted,withone
categoricalvariablebeingthebreakfastsize(small,medium,large)andtheotherbeinga
categoricalstresslevel(nostress,littlestress,moderatestress,oralotofstress,verysimilarto
the15scale).Thiscouldbebetterbecauseavisualrepresentationofthedatacouldbeabarplot
insteadofahistogramandnormalityofthedata(difficulttoachievewithasubjective
quantitativemeasurementlikestress)isnolongerarequiredconditionforthehypothesistest,
andisinsteadreplacedbyexpectedcountsconditionsthatcaneasilybemetbytheincreased
samplesizeimprovementmentionedearlier.Regardlessofthesefutureimprovements,this
experimentinthiscurrentstateaswellasthecollecteddatadidnotprovideenoughevidenceto
supportthateatingalargerbreakfastcanreducetheamountofstressbeforeanevent.


Sheth45

Appendix K - Peer Review Feedback Form


Sheth46

You might also like