Tải bản đầy đủ - 0 (trang)
Hack 71. Discover a New Species

Hack 71. Discover a New Species

Tải bản đầy đủ - 0trang

Forsomegroupofanimalstotechnicallybeaseparatespecies,

itmustshareauniquesetofbiologicalcharacteristicsthat

makeitdistinctfromsimilaranimals.Sure,animalswithinthe

samefamilyalllookalittledifferentfromeachother,butthen,

peoplelookalotdifferentfromeachotherandweareallone

species(myUncleFrankbeingperhapstheexceptionthat

provestherule).

Ifagroupofanimals,suchasDr.Cunningham'spossums,have

moreincommonwitheachotherthantheydowiththeother

creaturesintheirspecies,theymightbecandidatesfor

considerationasaspeciesintheirownright.Statisticscan

determinethat"morelikeeachotherandmoredifferentfrom

therestofthespeciesthanchancealonewouldproduce"point.

UsingCunningham'sdiscoveryasamodel,thereareafew

stepstofollowforyoutomakeyourowndiscovery.



Collectsomedata

ThispossumexistedinAustralianearpeopleformorethan200

yearsandnoonenoticed.Tobefair,itlookedanawfullotlike

theotherpossums,themostcommonofwhichwasthe

trichosuruscaninus,nowcalledtheshort-earedpossum.

Itwasassumedforsometimethattherewasreallyjustthis

onespeciesofthelittleguys.PartofDr.Cunningham'sjobwas

tocollectandorganizedescriptivedataforthewildlifearound

him.Consequently,hehadatonofveryspecificquantitative

descriptionsofvariouspossumpartseyes,ears,nose,and

throatandmeasurementsofotherphysicalcharacteristics.



Chooseastatisticalmethod

Cunningham'schoicewasatechniquesimilartofactoranalysis



butwithamoreimposingname:canonicalvariateanalysis.You

canuseanymethodthatusesthevariabilityinscorestocreate

distinctgroupings.Someofthosearediscussedinthis

booksuchasfactoranalysis,mentionedearlierinthishackbut

therearemanyotherproceduresthatwouldwork.



Ifyouarereallystatisticallysavvy,itwillhelpyoutoknowthat

canonicalvariateanalysisisfunctionallythesameasdiscriminant

analysisormultivariateanalysisofvariance(MANOVA),twoother

proceduresthatcreatelinearcompositesofvariableswiththegoalof

conceptuallydefiningtwoormoredistinctlydifferentgroups.



Cunninghamusedthisstatisticalproceduretoexaminethe

descriptivedataforthispresumablysinglespecies(youknow,

thesetrichosuruscaninusfellers)anddemonstratedthatthere

werelikelytwodifferentspecies.



Selectahypothesisandanalyzethedata

Statisticianstesthypotheses,soyoushouldbeginyouranalysis

withaguessaboutwhetherthereisorisnotadistinction

betweenthegroupsofparticipateswhosuppliedyourdata.

Intheexampleofourhero,Cunninghamassumedthatthere

weretwodifferentgroupsofcrittersthataccountedforthe

data.Then,theprocedure(usingacomputerforthe

calculations,ofcourse)identifiedwhichvariablesworkedbest

askeydistinguishingcharacteristicsbetweenthetheoretical

groups.



Thedifferencebetweenusingthistool,canonicalvariateanalysis,and

somethinglikeregressionisthat,whenusingvariablestomake

predictionsinregression,theresearcherhassomeknowndataabout



scoresofactualsubjects:which"group"theybelongto[Hack#13].

Here,theprocedureworksblindlywithoutknowingwhatthecorrect

answeris.Instead,itfindsgroupsthatcanbemadethemostdifferent

withthevariablesathand.



HerearethevariablesCunninghamused:

Headlength

Skullwidth

Eyesize

Earlength

Bodylength(fromtipofnosetotipofuncurledtail)

Taillength

Chestwidth

Footlength

Whileothervariableswereconsidered,Cunninghamchose

thesebecausetheywereeventuallyfoundtobemostimportant

indistinguishingonespeciesfromanotherandalsobecause

theywerecharacteristicsthatwouldprobablybeunaffectedby

environment.



Interpretresults



Thelaststepinanystatisticalanalysisistodescribeand

understandwhateveryoufound.Fordiscoveringspecies,you

needtobeabletodescribethatnewspeciesinenoughdetailto

differentiateitformother,similarspecies.

TheproceduresusedbyCunninghamidentifiedaseriesof

differentequationsthatweightedeachofthebiological

variablesdifferently,tofindthecombinationthatbestidentified

twoseparategroups.Theseequations(whichtheprocedure

labelsvariates)aresimilartoregressionequations,withthe

outcomeorcriterionvariabledeterminingwhichgroupa

possumbelongsto.

Here'sthesinglebestequationthataccountedforan

astonishing89percentofthevariabilityonthesecharacteristics

forallthepossumsinhisdatabase:

(headlengthx.44)+(skullwidthx.07)+(eyesizex.05)+

(earlengthx.82)+(bodylengthx.35)+(taillengthx.72)+

(chestwidthx.16)+(footlengthx.70)

I'veprovidedthestandardizedweightsfromthestudy,sowe

cancomparethemtoeachother.Thelargerweightsindicate

thepossumpartsthatdifferedthemostbetweenthe

mathematicallychosentwogroupsofpossums.

Inthisdata,youcouldfindtwogroupsofpossumsthatdiffered

themostbasedonearlength,taillength,andfootlength.The

amountofvariabilityexplainedwassolargethat,statistically,

Cunninghamconcludedthatthemathematicallyidentified

groupingswerereal.Thetwogroupsofpossumsfoundinthe

datawereactuallytwodifferentspeciesofpossum,andthe

speciescouldbedefinedbytheirearlengthandacoupleof

othervariables.Thelargertheweightsintheequationshown

earlier,themorethetwospeciesdifferedonthesebodyparts.



TwoPossumSpecies



Table6-20showstheofficialdescriptionsofthetwopossum

speciesfirstidentifiedassuchbyourstatisticianandhis

mathematics.Noticethenamesareevenbasedonthekey

predictorsfoundinthestatisticalanalysis!

TableTwocommonAustralianpossums





Commonname

Habitat

Ears

Feet

Head

Tail



trichosuruscaninus

Short-earedpossum

Livesinthenorth

Shorterears

Smallerfeet

Biggerhead

Longertail



trichosuruscunninghamii

Mountainbrushtailpossum

Livesinthesouth

Longerears

Largerfeet

Smallerhead

Smallertail



So,startcollectingyourowndataonthoseodd,stinkybugs

youfindonyourscreendoorandyouarewellonyourwayto

greatnessandimmortality.Isthereonespeciesofstinkbugor

two?Youtellme.



SeeAlso

Ifirstlearnedaboutthisapproachtoidentifyingspeciesin

thisfinearticle:Hall,P.(2003).Chance,16,1.



Hack72.FeelConnected



Theconceptof"sixdegreesofseparation"ismorethan

justaNewAgemetaphorforcommunityorapartygame

involvingtheactorKevinBacon.Ifyouwanttoactually

testtheideathatweallknowsomeonewhoknows

everybodyelse,findouthowcloselylinkedyoureallyare

toeveryone.

Iknowaguywhoknewaguywhousedtoworkforthe

PresidentoftheUnitedStates.Smallworld,eh?I'mnotsaying

Ihavegreatconnections,butIamjusttwohandshakesaway

fromtheleaderofthefreeworld.Beforeyougettooimpressed,

youshouldknowthatyouprobablyarejustafewlinksaway

fromalmostanybodyintheworld.

Itisprobablytruethatanytwopeoplearewithinsixdegreesof

separation,andthatmagicandoft-quotednumberof6is

actuallytakenfromarealscientificstudy!Herearesomeclever

researchmethodstoletyourevealtheinvisibleconnections

thatuniteusall,oratleastlinkyoutothatpersonontheother

sideofthecocktailparty.



SixDegreesofSeparation

ThereisaplaycalledSixDegreesofSeparationbyJohnGuare

andamoviebasedonthatplaystarringWillSmith.Thereis

alsoapopularpartytriviagame,sometimescalledSixDegrees

ofKevinBacon,thatattemptstolinkanyactororactress

throughaseriesofmoviesandotherperformersuntilthey

shareaconnectionwithactorKevinBacon.



Thephraseandconceptcomefromastudythatconsideredthe

small-worldproblem.Haveyoueverbeenatapartyorbeen

chattingwithastrangeratacoffeeshopanddiscoveredthat

youbothknowthesameperson?SocialpsychologistStanley

Milgramwascuriousaboutthisphenomenoninthelate1960s

(whentherewerealotmorecocktailpartiesthanthereare

now).Howmuchoverlapwasthereinsocialnetworks?Ifwe

couldallgettogetherandlisteveryoneweknow,wouldthere

alwaysbesomeconnection?Probably,eventually,aswe

exploredfurtherandfurtheroutofthecenterofourwebof

acquaintances,wewouldfindsomeconnectionwithalmost

everyone.Buthowmanylinkswouldittake?

Justonedegreeofseparationmeansweallknoweveryone.

Well,Idon'tknowyou(nooffense),soweknowthatoneistoo

fewlinkstoconnecteveryone.Aretherejusttwodegreesof

separation?Ifwedon'tknoweachother,maybewehavea

friendincommon?

Thequestion,therefore,ishowmanydegreesofseparationare

therebetweenyouandanyoneelse?Togettheanswer,doabig

studyorasmallstudyusingthemethodsinthishack.



DoingaBigStudy

Howcouldonestudytheproblemofwhetherweactuallylivein

asmallworld?Thebestwayistoduplicatethemethodsused

byStanleyMilgram.



Chooseatarget

Milgramstartedbypickingsomeoneheknewwhoworkedin

Boston,Massachusetts,whereMilgramlived.Itwasn'tKevin

Bacon,butastockbrokerwhoagreedtoactasthetarget,the

finalendofachainthatMilgramhopedtobuild.Youcouldpick



yourbestfriendoryourschoolprincipaloryourUniversity's

president.Yougottaasktheirpermissionfirst,though

(somethingaboutethics).



Recruitparticipates

Milgramthenrandomlysampledfromtwocommunities:Boston

andOmaha,Nebraska.Thissamplingschemewasmeantto

representthetwoextremesoflikelihoodthatanyonewould

knowthetarget.Startwithpeopleclosebyandpeoplefar

away,andtheaverageoftheirdatashouldbefairly

representativeofthepopulation.Milgramused300randomly

chosenrecruits.Youshoulduseasmanyasyoucanaffordor

havetimefor.



Trainparticipates

Milgramsentapacketinthemailtoeachrecruit.Thepacket

containedinstructionsdescribingthestudyandaletterforthe

Bostonbroker.Theywereaskedtodeliverthelettertoourguy,

butonlyiftheyknewhimpersonally.Iftheydidnotknowhim

personally,theywereaskedtorecordsomeinformation,such

astheirname,andsendthepacketontosomeonewhothey

didknowwhotheythoughtmighthaveabetterchanceof

knowinghim.Thosenextpeopleinthechainreceivedthesame

packetwiththeinstructionsandtheletter.Theymighthave

sentittothebrokeriftheyknewhim,orsentitontoathird

linkinthechain,andsoon.

Inyourownstudy,makesuretowritetheinstructionsclearly

andsimply,and,thesedays,youmightexplainthatthisis

legitimateresearch,notacommercialsolicitationandnota

chainletter(thoughitliterallyis,Iguess),andallthe

disclaimersyouthinkwillhelp.Youshouldalsoincludecontact

informationforyouifanyonehasanyquestionsaboutthe



legitimacyoftheproject.



Collectandanalyzetheresults

Afterareasonableamountoftime,checkwithyourtargetand

gatherallthelettersreceived.Oneachletter,countthenumber

ofnamesthatformthechain.Averageallthedifferentlengths

ofchainstodeterminethetypicalnumberofconnections.Find

thesmallestnumbernecessarytoincludeeventhelongest

chain,andyouhavethemaximumdistance.

TheBostontargetinMilgram'sstudyeventuallyreceivedabout

100letters.Ofthose,theaveragenumberoflinkswassixthus,

theoriginofthenumbersixin"sixdegreesofseparation."

Notice,however,thatnotalllettersarrived,sowedon'tknow

fromthisonestudythatsixisreallytherightnumber.The

studyalsotookplaceintheU.S.only,notworldwide,so

granderviewsoftherebeingonlyafewdegreesofseparation

betweenanytwopeopleonthewholeplanetarephilosophically

based,notempiricallyderived.



TheresponseratethatMilgramenjoyedwasveryhigh,consideringthe

complicatedrequestsmadeofparticipants.Thisisnotsurprising,

becauseMilgramknewsomethingaboutobedience.StanleyMilgramis

probablybetterknownforanothercleverstudywithmoredisturbing

resultsheconductedsomeyearsbeforehissmallworldstudy.Withhis

obediencestudiesoftheearly1960s,Milgramdemonstratedthatwhen

peopleofauthority(suchasresearchassistantsinlabcoats)askstudy

participantstodosomethingthatmakesthemuncomfortable,suchas

administering(orbelievingthattheyareadministering)anelectric

shocktoanotherresearchsubject,asurprisingnumberofpeoplewill

doit.Hisresearchledtomuchinsightastowhypeoplemight"obey

orders"eveniftheydisagreewiththem.



Twomorerecentstudieshaveconfirmedthattheaverage

numberofconnectionsbetweenpeopleinsocialnetworksis

aboutsixorevenalittleless.



DoingaSmallStudy

Thereareacoupleofwaystousethesemethodsthatdon't

takequiteasmuchwork.Thegoaloftheactivitycouldbe

scientificorjustpartyfun.



Milgramviaemail

DuplicatetheMilgramstudy,butusetheconvenienceofemail.

Here,thequestionwouldbehowmanylinksbetweenpeople

usingtheiremailaddresses.Emailiseasiertoworkwiththan

snailmailandisvirtuallycost-free.

Ofcourse,choosingrecruitsthroughemailisprobablymore

difficult.Itishardtochooseemailaddressesrandomly,because

thereisn'tabigphonebook-typelisttosamplefrom.Also,your

emailrequestsmightquicklybemistaken(?)forspamand

ignored.Bytheway,becauseyourresearchinterestis

legitimate,youshouldn'thavetoworryaboutviolatingany

Internetprotocols.



Throwaparty

Whenhostingalargeparty(Milgramwouldhaveloveditifyou

usedacocktailparty,theinspirationforhisoriginalstudy),

handoutsuppliestoyourguests.Givethemeachalargeindex

cardandapen.Atthebottomofeachcard,listthenameofa

guestattheparty.Ifguestsdon'tknowthepersonlistedbelow,

theyshouldsigntheirnameatthetopofthecardandhandit



tosomeoneelsewhotheythinkmightknowtheperson.

Theprocessshouldcontinue,justasintheMilgramstudy,until

thecardsreachthepersonwhoisnamedonthebottom.That

personthenturnsthecardin.Attheendoftheparty,youcan

analyzethedataandprovetoyourgueststhattheyallreally

knoweachother.



JustDoingtheMath

Evenwithoutscientificstudies,however,aquickmathematical

analysismightconvinceyouthatthenumberofpeoplebetween

youandanyoneelseisafairlylownumber.Howmanypeople

doyouknowbytheirfirstnames?100?200?Let'ssayitis

about100.Theyeachknowabout100peoplebytheirfirst

names,too,presumably,soyouarealreadyconnectedto

10,000peoplethroughjusttwodegreesofseparation.

(Actually,10,100,intotal,countingthe100peoplewhoare

withinonedegreeofyou.)Itwouldn'ttaketoomanydegrees

beforeyouareconnectedtoawholelotofpeople,asshownin

Table6-21.

TableDegreesofseparationandcorrespondingconnections

Degreesofseparation

Connections

1

100

2

10,000

3

1,000,000

4

100,000,000

5

10,000,000,000



Infact,withjustfivedegreesofseparation,youshouldbe

connectedto10billionpeople,morethanthereareonearth!

So,why,inreality,areagreaternumberofconnectionsneeded



Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Hack 71. Discover a New Species

Tải bản đầy đủ ngay(0 tr)

×