Sunteți pe pagina 1din 10

Stat250GundersonLectureNotes

3:Probability

Chancefavorspreparedminds.LouisPasteur

Manydecisionsthatwemakeinvolveuncertaintyandtheevaluationofprobabilities.

InterpretationsofProbability

Example:Rollafairdiepossibleoutcomes={1,2,3,4,5,6}

Beforeyourollthediedoyouknowwhichonewilloccur?No
Whatistheprobabilitythattheoutcomewillbea4?1/6=P(4)Why?

AfewwaystothinkaboutPROBABILITY:

(1)PersonalorSubjectiveProbability
P(A)=thedegreetowhichagivenindividualbelievesthattheeventAwillhappen.

(2) Longtermrelativefrequency
P(A)=proportionoftimesAoccursiftherandomexperiment(circumstance)isrepeated
many,manytimes.

10balls:3blueand7white;Oneballwillbeselectedatrandom.
(3)BasketModel
WhatisP(blue)?__3/10__
P(A)=proportionofballs
inthebasketthathave
anAonthem.
Note:eachtimeIdotheexperiment,
theselectedballiseitherwhiteorblue;
onceIlook,thereisnomoreprobability)

Note:
AprobabilitystatementISNOTastatementaboutINDIVIDUALS.

ItISastatementaboutthepopulation/thebasketofballs.

DiscoverBasicRulesforFindingProbabilitythroughanExample
Thereisalotyoucanlearnaboutprobability.Onebasicruletoalwayskeepinmindisthatthe
probabilityofanyoutcomeisalwaysbetween0and1.Now,thereareentirecoursesdevoted
just to studying probability. But this is a Statistics class. So rather than start with a list of
definitionsandformulasforfindingprobabilities,letsjustdoitthroughanexamplesoyoucan
seewhatideasaboutprobabilityweneedtoknowfordoingstatistics.

35

Example:ShoppingOnline
ManyInternetusersshoponline.Considerapopulationof1000customersthatshoppedonline
ataparticularwebsiteduringthepastholidayseasonandtheirresultsregardingwhetherornot
theyweresatisfiedwiththeexperienceandwhetherornottheyreceivedtheproductsontime.
Theseresultsaresummarizedbelowintableform.Usingtheideaofprobabilityasaproportion,
tryansweringthefollowingquestions.

OnTime
NotOnTime Total
Satisfied
800
20
820
NotSatisfied
80
100
180
Total
880
120
1000

MaydecidetonotincludetheprobabilitynotationP()thefirsttimethrough.
a. Whatistheprobabilitythatarandomlyselectedcustomerwassatisfiedwiththeexperience?
Thinkaboutprobabilityasproportion,so820/1000=0.82or82%
ThiswouldberepresentedbyP(Satisified).

b. What is the probability that a randomly selected customer was not satisfied with the
experience?If82%weresatisfied,then10.82=0.18or18%werenotsatisfied.Wouldbe
representedbyP(NotSatisfied),andyoujustusedwhatwecallthecomplementrule.

c. Whatistheprobabilitythatarandomlyselectedcustomerwasbothsatisfiedandreceived
theproductontime?Againfindtheproportionofcustomersthathavebothofthesetraits
800/1000=0.80or80%.ThiswouldberepresentedbyP(SatandOT).

d. Whatistheprobabilitythatarandomlyselectedcustomerwaseithersatisfiedorreceived
theproductontime?HerewehaveanOR.Soweneedtheproportionofcustomersthat
wereeithersatisfiedorreceiveditontime.Startwithhowmanyweresatisfied(820)then
letsaddinthosethatreceiveditontime(880),butindoingso,weincludedthosethat
werebothsatisfiedandreceivedontime(800)twice!Soweneedtocorrectforthatdouble
counting (820 + 880 800) / 100 = 900/1000 = 0.90 or 90%. P(satisfied or on
time)=P(satisfied)+P(ontime)P(satisfiedandontime)=0.82+0.880.80=0.90

e. Given that a customer did receive the product on time, what is the probability that the
customerwassatisfiedwiththeexperience?Youcanjustusethetabletologicallyfigureit
out.Weshouldonlylookatthosecustomersthatreceiveditontimetherewere880.
Outofthese880,letsfindtheprobability(theproportion)theyweresatisfied800outof
880or0.901(90.1%).Whatyoujustfoundwasaconditionalprobability:P(Satisfied|on
time)wherethelinemeansGivenorConditionalonoramongthose

f. Giventhatacustomerdidnotreceivetheproductontime,whatistheprobabilitythatthe
customerwassatisfiedwiththeexperience?
Outofthe120thatdidnotreceiveontime,theproportionthatweresatisfied20outof
120or0.167(16.7%).ThisconditionalprobabilityisP(Satisfied|Notontime);notehow
muchloweritiscomparedtopart(a)andpart(e).

Note:Westatedtheabove1000customersrepresentedapopulation.Ifresultswerebasedonasamplethatis
representative of a larger population, then the observed sample proportions would be used as approximate
probabilitiesforarandomlyselectedpersonfromthelargerpopulation.

36

Greatjob!Youjustcomputedprobabilitiesusingmanyofthebasicprobabilityrulesorformulas
summarizedbelowandalsofoundinyourtextbook.

P( AC ) 1 P( A)
Complementrule

Additionrule

P( A or B) P( A) P(B) P( A and B)

Multiplicationrule

P( A and B) P( A)P(B | A)

ConditionalProbability P ( A | B )

P ( A and B )

P( B)

Youdidnotneedtheformulasthemselvesbutinsteadusedintuitionandapproachingitassome
typeofproportion.Letsseehowyourintuitionandtheaboveformulasreallydoconnect.

InpartbyoufoundtheprobabilityofNOTbeingsatisfied,whichisthecomplementofthe
eventbeingsatisfied,sotheanswertopartbisthecomplementoftheprobabilityyoufound
inparta.

Inpartc,therewasakeywordofANDinthe
question being asked. The AND is just the
intersection, or the overlapping part; the
outcomesthatareincommon.Thepictureat
the right show an intersection between the
eventAandB.Inatable,thecountsthatarein
themiddlearetheANDcounts;therewere
800(outofthe1000customers)thatwereboth
satisfiedANDreceivedtheproductontime.Thereisamultiplicationformulaaboveforfinding
probabilitiesoftheANDorintersectionoftwoevents,butwedidnotevenneedtoapplyit;asa
tablepresentationofcountsprovidesANDcountsdirectly.

Inpartd,therewasakeywordofORinthequestionbeingasked.TheORisunion,the
outcomesthatareineitheroneortheother(includingthosethatareinboth).Thepictureat
the right show an union between the event A and B. Notice that if you start with all of the
outcomesthatareinAandthenaddalloftheoutcomesthatareinB,youhavedoublecounted
theoutcomesthatareintheoverlap.Sotheadditionformulaaboveshowsyouneedtosubtract
offtheintersectingprobabilityoncetocorrect
for the double counting. From the table, you
couldeitheradduptheseparatecountsof800
+ 20 + 80; or start with the 820 that were
satisfied and add the 880 that received it on
time and then subtract the 800 that were in
bothsets;togetthe900inallthatwereeither
satisfiedorreceivedtheproductontime.

37

Finally,partseandfwerebothconditionalprobabilities.Inparteyouwerefirsttoldtoconsider
onlythe880customersthatreceivedtheproductontime,andoutofthesefindtheprobability
(or proportion) that were satisfied. There were 800 out of the 880 that were satisfied. The
picturebelowshowstheideaofaconditionalprobabilityformulaaboveforP(A|B),readasthe
probabilityofAgivenBhasoccurred.IfweknowBhasoccurred,thenonlylookatthoseitems
intheeventB.TheeventB,shadedattheright,isournewbase(andthusisinthedenominator
oftheformula).NowoutofthoseitemsinB,
wewanttofindtheprobabilityofA.Theonly
itemsinAthatareonthesetBarethoseinthe
overlap or intersection. So the conditional
formulaaboveshowsyoucountupthoseinthe
AandBanddivideitbythebaseofB.

TryIt!Gobacktopartsatofandaddthecorrespondingshorthandprobabilitynotationofwhat
youactuallyfound;e.g.P(satisfied),P(satisfied|ontime)nexttoeachanswer.

Nowthereareacoupleofusefulsituationsthatcanmakecomputingprobabilitieseasier.

Definition:
TwoeventsA,BareMutuallyExclusive(orDisjoint)if...theydonotcontainany
ofthesameoutcomes.Sotheirintersectionisempty.
Wecaneasilypicturedisjointeventsbecausethedefinitionisapropertyaboutthesets
themselves.

IfA,Baredisjoint,thenP(AandB)=0.Iftherearenoitemsintheoverlappingpart,thenmanof
theprobabilityresultswillsimplify.Forexample,theadditionalrulefordisjointeventsP(AorB)
=P(A)+P(B).

38

Anotherimportantsituationinstatisticsoccurswhenthetwoeventsturnouttobeindependent.

Definition:
TwoeventsA,Baresaidtobeindependentifknowingthatonewilloccur
(orhasoccurred)doesnotchangetheprobabilitythattheotheroccurs.
InnotationthiscanbeexpressedasP(A|B)=P(A).

ThisexpressionP(A|B)=P(A)tellsusthatknowingtheeventBoccurreddoesnotchangethe
probabilityoftheeventAhappening.

Nowitworkstheotherwayaroundtoo,ifAandBareindependentevents,thenP(B|A)=P(B).

As a result of this independence definition, we could show that the multiplication rule for
independenteventsreducestoP(AandB)=P(A)P(B).

Finally,thisrulecanalsobeextended.IfthreeeventsA,B,CareallindependentthenP(AandB
andC)=P(A)P(B)P(C).

Soletsapplythesetwonewconceptstoouronlineshoppingexample.

BacktotheShoppingOnlineExample

Belowareresultsforapopulationof1000customersthatshoppedonlineataparticularwebsite
during the past holiday season. Recorded was whether or not they were satisfied with the
experienceandwhetherornottheyreceivedtheproductsontime.

OnTime
NotOnTime Total
Satisfied
800
20
820
NotSatisfied
80
100
180
Total
880
120
1000

g. Arebeingsatisfiedwiththeexperienceandreceivingtheproductontimemutuallyexclusive
(disjoint)?Providesupportforyouranswer.
i.e.arethereanycustomersintheintersection,thatwerebothsatisfiedandreceivediton
time? Since there are 800 customers in the intersection, these two events are NOT
MUTUALLYEXCLUSIVE(thatis,NOTDISJOINT).

h. Are being satisfied with the experience and receiving the product on time statistically
independent?Providesupportforyouranswer.
Hint:gobackandcompareyouranswerstopartsa,e,andf.
From(a)wehaveP(satisfied)=0.82;from(b)wehaveP(satisfied|ontime)=0.901;and
frompart(f)wehaveP(satisfied|notontime)=0.167.Sinceknowingtheproductwas
receivedontimedidincreasetheprobabilityofthecustomerbeingsatisfied(from0.82to
0.901),thetwoeventsareNOTINDEPENDENT.Note:wecouldalsocheckifP(Satisfiedand
OnTime)=P(Satistifed)P(OnTime).Thisisthemultiplicationruleforindependentevents.
Since0.80doesnotequal(0.82)(0.88)=0.7216,againthesameconclusionisreached
notindependent.

39

TryIt!ElderlyPeople

ElderlyPeople

Supposethatinacertaincountry,10%ofthe
elderly people have diabetes. It is also
known that 30% of the elderly people are
living below poverty level and 5% of the
elderly population falls into both of these
categories.

diabetes

below
poverty
0.05

Attherightisadiagramfortheseevents.
Dotheprobabilitiesmakesensetoyou?

0.05

0.25

diabetes & below poverty

Neither diabetes
nor below poverty
0.65

a. Whatistheprobabilitythatarandomlyselectedelderlypersonisnotdiabetic?

P(notdiab)=1P(diab)=10.10=0.90or90%

b. Whatistheprobabilitythatarandomlyselectedelderlypersoniseitherdiabeticorliving
belowpovertylevel?

P(diaborBP)=P(diab)+P(BP)P(diabandBP)=0.10+0.300.05=0.35or35%

c. Givenarandomlyselectedelderlypersonislivingbelowpovertylevel,whatistheprobability
thatheorshehasdiabetes?

P(diab|BP)=P(diabandBP)=0.05=0.167

P(BP)0.30

d. Sinceknowinganelderlypersonlivesbelowthepovertylevel(circleone)

DOES DOESNOTchangetheprobabilitythattheyarediabetic,thetwo

eventsoflivingbelowthepovertylevelandbeingdiabetic(circleone)

ARE ARENOTindependent.

40

Inthenextexample,youarenotaskedtodetermineiftwoeventsareindependent,butrather
putindependencetouse.

TryIt!BloodType
About1/3ofalladultsintheUnitedStateshavetypeO+blood.Supposethreeadultswillbe
randomlyselected.
Hint:randomlyselectedimpliestheresultsshouldbe_independent(&identicallydistributed)_.

WhatistheprobabilitythatthefirstselectedadultwillhavetypeO+blood?
P(firsthasO+)=1/3=0.333

WhatistheprobabilitythatthesecondselectedadultwillhavetypeO+blood?
P(secondhasO+)=1/3=0.333

WhatistheprobabilitythatallthreewillhavetypeO+blood?
P(1sthasO+and2ndhasO+and3rdhasO+)
=P(1sthasO+)P(2ndhasO+)P(3rdhasO+)=(1/3)(1/3)(1/3)=0.037

WhatistheprobabilitythatnoneofthethreewillhavetypeO+blood?
P(1stnotO+and2ndnotO+and3rdnotO+)=(2/3)(2/3)(2/3)=0.296

WhatistheprobabilitythatatleastonewillhavetypeO+blood?
P(atleast1hasO+)=1P(nonehaveO+)=10.296=0.704

Somefinalnotes
I. SamplingwithandwithoutReplacement

Definitions:
Asampleisdrawnwithreplacementifindividualsarereturnedtotheeligiblepool
foreachselection.Asampleisdrawnwithoutreplacementifsampledindividuals
arenoteligibleforsubsequentselection.

Ifsamplingisdonewithreplacement,theExtensionofRule3bholds.Ifsamplingisdonewithout
replacement, probability calculations can be more complicated because the probabilities of
possibleoutcomesatanyspecifictimeinthesequenceareconditionalonpreviousoutcomes.

Ifasampleisdrawnfromaverylargepopulation,thedistinctionbetweensamplingwithand
without replacement becomes unimportant. In most polls, individuals are drawn without
replacement,buttheanalysisoftheresultsisdoneasiftheyweredrawnwithreplacement.The
consequencesofmakingthissimplifyingassumptionarenegligible.

41

II. Sometimesstudentsconfusethemutuallyexclusivewithindependence.

Checkthedefinitions.
Thedefinitionfortwoeventstobedisjoint(mutuallyexclusive)wasbasedonaSET
property.
ThedefinitionfortwoeventstobeindependentisbasedonaPROBABILITYproperty.

Youneedtocheckifthesedefinitionsholdwhenaskedtoassessiftwoeventsaredisjoint,
oriftwoeventsareindependent.

MutuallyExclusive Independence

IftwoeventsaremutuallyexclusivethenweknowthatP(AandB)=0.
ThisalsoimpliesthatP(A|B)isequalto0(ifthetwoeventsaredisjointandBdidoccur,
thenthechanceofAoccurringis0).
SoP(A|B)(whichis0)willnotbeequaltoP(A)iftheeventsaredisjoint.
Considerapersonsnaturaleyecolorandletthetwoeventsbeeyecolorisbrownandeyecolorisgreen.
For a given population that 25% have brown eyes, 16% have green eyes, and remaining 59% of the
populationhavesomeothercolorofeyes.P(Brown)=0.25,P(Green)=0.16,P(other)=0.59(i)Arethetwoevents
greeneyesandbrowneyesmutuallyexclusive?Yes,theintersectionBrownEyesANDGreenEyesis
empty. (ii) Are the two events green eyes and brown eyes independent? Does P(Brown|Green) =
P(Brown)?;Well,0doesnotequal0.25,sotheyareNOTindependent.

III.Probabilityrulessummary
Belowisasummaryofthekeyprobabilityresultsyouneedtounderstandandbeabletouse.

Complementrule
P( A C ) 1 P ( A)

MutuallyExclusive(disjoint)Events:
TheeventsA,BaredisjointifAandBistheemptyset.
Thus,P(AandB)=0.

AdditionRule(general) P( A or B) P( A) P(B) P( A and B)


IfA,Baredisjoint,wehave P( A or B) P( A) P(B)
P( A and B)
P( A | B)

ConditionalProbability(general)
P( B)
IndependentEvents:
TheeventsA,Bareindependentif P( A | B) P( A)
Equivalently,theeventsA,Bareindependent
if P( A and B) P( A)P(B)

TheStats250formulacardprovidesamoreextensivelist,butremember,youmaynotneed
themasyoudiscoveredinyourfirstprobabilityexamplewiththeonlinecustomers.

42

AdditionalNotes
Aplacetojotdownquestionsyoumayhaveandaskduringofficehours,takeafewextranotes,write
outanextraproblemorsummarycompletedinlecture,createyourownsummaryabouttheseconcepts.

43

44

S-ar putea să vă placă și