stat 13, intro. to statistical methods for the life and health …frederic/13/sum17/day09.pdf ·...

76
Stat 13, Intro. to Statistical Methods for the Life and Health Sciences. 0. Hand in hw4. 1. Calculating correlation. 2. Slope of the regression line. 3. Goodness of fit. 4. Common problems with regression. a. Inferring causation. b. Extrapolation. c. Curvature. 5. Testing significance of correlation or slope. 6. ANOVA and F-test, ch9. 7. Review and examples. The final is Thu in class and will be on ch 1-7, 10, and at most 1 question on ch9. Bring a PENCIL and CALCULATOR and any books or notes you want. No computers. http://www.stat.ucla.edu/~frederic/13/sum17 . 1

Upload: others

Post on 18-Mar-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Stat 13, Intro. to Statistical Methods for the Life and Health Sciences.

0.Handinhw4.1.Calculatingcorrelation.2.Slopeoftheregressionline.3.Goodnessoffit.4.Commonproblemswithregression.

a.Inferringcausation.b.Extrapolation.c.Curvature.

5.Testingsignificanceofcorrelationorslope.6.ANOVAandF-test,ch9.7.Reviewandexamples.

ThefinalisThuinclassandwillbeonch 1-7,10,andatmost1questiononch9.BringaPENCILandCALCULATORandanybooksornotesyouwant.Nocomputers.http://www.stat.ucla.edu/~frederic/13/sum17.

1

Page 2: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

1.Calculatingcorrelation,r.

ρ =rho=correlationofthepopulation.SupposethereareNpeopleinthepopulation,X=temperature,Y=heartrate,themeanandsd oftempinthepop.areµ" and𝜎" ,andthepop.meanandsd ofheartrateareµ$and𝜎$.

ρ =*+∑ "-./0

10+23*

$-./414

.

Givenasampleofsizen,weestimateρ using

r= *5.*

∑"-.060

523*

$-.464

.

ThisisinAppendixA.

2

Page 3: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

2.Slopeofregressionline.• Suppose𝑦8 =a+bx istheregressionline.

• Theslopeboftheregressionlineisb=r6460.

Thisisusuallythethingofprimaryinteresttointerpret,asthepredictedincreaseinyforeveryunitincreaseinx.• Bewareofassumingcausationthough,esp.withobservationalstudies.Bewaryofextrapolationtoo.

• Theintercepta=$ - b " .

• TheSDoftheresidualsis 1 − 𝑟<� 𝑠$.Thisisagoodestimateofhowmuchtheregression

predictionswilltypicallybeoffby.

Page 4: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

3.Howwelldoesthelinefit?• 𝑟< isameasureoffit.Itindicatestheamountofscatteraroundthebestfittingline.

• 1 − 𝑟<� 𝑠$isusefulasameasureofhowfaroffpredictionswouldhavebeenonaverage.• Residualplotscanindicatecurvature,outliers,orheteroskedasticity.

• Notethatregressionresidualshavemeanzero,whetherthelinefitswellorpoorly.

-2 -1 0 1 2 3 4 5

-4-2

02

46

8

x

y

-2 -1 0 1 2 3 4 5-10

-8-6

-4-2

02

x

residual

Page 5: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

• Heteroskedasticity:whenthevariabilityinyisnotconstantasxvaries.

Page 6: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

4.Commonproblemswithregression.• a.Correlationisnotcausation.Especiallywithobservationaldata.

Page 7: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.

Page 8: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.Holmes and Willett (2004) reviewed all prospective studies on fat consumption and breast cancer with at least 200 cases of breast cancer. "Not one study reported a significant positive association with total fat intake.... Overall, no association was observed between intake of total, saturated, monounsaturated, or polyunsaturated fat and risk for breast cancer."

They also state "The dietary fat hypothesis is largely based on the observation that national per capita fat consumption is highly correlated with breast cancer mortality rates. However, per capita fat consumption is highly correlated with economic development. Also, low parity and late age at first birth, greater body fat, and lower levels of physical activity are more prevalent in Western countries, and would be expected to confound the association with dietary fat."

Page 9: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• b.Extrapolation.

Page 10: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• b.Extrapolation.• Oftenresearchersextrapolatefromhighdosestolow.

Page 11: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• b.Extrapolation.Therelationshipcanbenonlinearthough.Researchersalsooftenextrapolatefromanimalstohumans.

Zaichkina etal.(2004)onhamsters

Page 12: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• c.Curvature.Thebestfittinglinemightfitpoorly.Portetal.(2005).

Page 13: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• c.Curvature.Thebestfittinglinemightfitpoorly.Wongetal.(2011).

Page 14: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Commonproblemswithregression.• d.Statisticalsignificance.Couldtheobservedcorrelationjustbeduetochancealone?

Page 15: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

5.InferencefortheRegressionSlope:Theory-BasedApproachSection10.5

Page 16: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Dostudentswhospendmoretimeinnon-academicactivitiestendtohavelowerGPAs?Example10.4

Page 17: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Dostudentswhospendmoretimeinnon-academicactivitiestendtohavelowerGPAs?

• Thesubjectswere34undergraduatestudentsfromtheUniversityofMinnesota.

• Theywereaskedquestionsabouthowmuchtimetheyspentinactivitieslikework,watchingTV,exercising,non-academiccomputeruse,etc.aswellaswhattheircurrentGPAwas.

• WearegoingtotesttoseeifthereisanegativeassociationbetweenthenumberofhoursperweekspentonnonacademicactivitiesandGPA.

Page 18: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Hypotheses• NullHypothesis:ThereisnoassociationbetweenthenumberofhoursstudentsspendonnonacademicactivitiesandstudentGPAinthepopulation.

• AlternativeHypothesis:ThereisanegativeassociationbetweenthenumberofhoursstudentsspendonnonacademicactivitiesandstudentGPAinthepopulation.

Page 19: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

DescriptiveStatistics• GPAU = 3.60 − 0.0059(nonacademichours).• Istheslopsignificantlydifferentfrom0?

Page 20: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ShuffletoDevelopNullDistribution

• Wearegoingtoshufflejustaswedidwithcorrelationtodevelopanulldistribution.

• Theonlydifferenceisthatwewillbecalculatingtheslopeeachtimeandusingthatasourstatistic.

• atestofassociationbasedonslopeisequivalenttoatestofassociationbasedonacorrelationcoefficient.

Page 21: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Betavs Rho• Testingtheslopeoftheregressionlineisequivalenttotestingthecorrelation (samep-value,butobviouslydifferentconfidenceintervalssincethestatisticsaredifferent)

• Hencethesehypothesesareequivalent.• Ho:β =0Ha:β <0(Slope)• Ho:ρ =0Ha:ρ < 0(Correlation)

• Sampleslope(b)Population(β:beta)• Samplecorrelation(r)Population(ρ:rho)

• Whenwedothetheorybasedtest,wewillbeusingthet-statisticwhichcanbecalculatedfromeithertheslopeorcorrelation.

Page 22: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Introduction• Ournulldistributionsareagainbell-shapedandcenteredat0(foreithercorrelationorslopeasourstatistic).

The book on p549 finds a p value of 3.3% by simulation.

Page 23: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ValidityConditions• Undercertainconditions,theory-basedinferenceforcorrelationorslopeoftheregressionlineuset-distributions.

• Wecouldusesimulationsorthetheory-basedmethodsfortheslopeoftheregressionline.

• Wewouldgetthesamep-valueifweusedcorrelationasourstatistic.

Page 24: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

PredictingHeartRatefromBodyTemperatureExample10.5A

Page 25: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp• Earlierwelookedattherelationshipbetweenheartrateandbodytemperaturewith130healthyadults

• PredictedHeartRate = −166.3 + 2.44 Temp• r=0.257

Page 26: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp

• Wetestedtoseeifwehadconvincingevidencethatthereisapositiveassociationbetweenheartrateandbodytemperatureinthepopulationusingasimulation-basedapproach.(Wewillmakeit2-sidedthistime.)

• NullHypothesis:Thereisnoassociationbetweenheartrateandbodytemperatureinthepopulation. β=0

• AlternativeHypothesis:Thereisanassociationbetweenheartrateandbodytemperatureinthepopulation. β≠0

Page 27: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp

Wegetaverysmallp-value(0.0036).Anythingasextremeasourobservedslopeof2.44happeningbychanceisveryrare

Page 28: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp

• Wecanalsoapproximatea95%confidenceintervalobservedstatistic+ multiplierx SE2.44± 1.96x 0.842=0.790to4.09

• Whatdoesthismean?We’re95%confidentthat,inthepopulationofhealthyadults,each1° increaseinbodytempisassociatedwithanincreaseinheartrateofbetween0.790to4.09beatsperminute.

Page 29: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp

• Thetheory-basedapproachshouldworkwellsincethedistributionoftheslopeshasanicebellshape

• Alsocheckthescatterplot

Page 30: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

HeartRateandBodyTemp

• Wewillusethet-statistictogetourtheory-basedp-value.

• Wewillfindatheory-basedconfidenceintervalfortheslope.

• Onp554,thebooknotestheformulat= cdefgheg

�.

• Herethetstatisticis2.97.• Thep-valueis0.36%.Sothecorrelationisstatisticallysignificantlygreaterthanzero.

Page 31: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

SmokingandDrinkingExample10.5B

Page 32: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ValidityConditionsRememberourvalidityconditionsfortheory-basedinferenceforslopeoftheregressionequation.

1. Thescatterplotshouldfollowalineartrend.2. Thereshouldbeapproximatelythesamenumber

ofpointsaboveandbelowtheregressionline(symmetry).

3. Thevariabilityofverticalslicesofthepointsshouldbesimilar.Thisiscalledhomoskedasticity.

Page 33: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ValidityConditions• Let’slookatsomescatterplotsthatdonotmeettherequirements.

Page 34: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

SmokingandDrinkingTherelationshipbetweennumberofdrinksandcigarettesperweekforarandomsampleofstudentsatHopeCollege.

Thedotat(0,0)represents524students

Aretheconditionsmet?Hardtosay.Thebooksaysno.

Page 35: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

SmokingandDrinking• Whentheconditionsarenotmet,applyingsimulation-basedinferenceispreferabletotheory-basedt-testsandCIs.

Page 36: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ValidityConditions

• Whatdoyoudowhenvalidityconditionsaren’tmetfortheory-basedinference?• Usethesimulated-basedapproach.

• Anotherstrategyisto“transform”thedataonadifferentscalesoconditionsaremet.• Thelogarithmicscaleiscommon.

• Onecanalsofitadifferentcurve,notnecessarilyaline.

Page 37: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

6.ANOVAandF-test.

Section9.2

Page 38: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ANOVA• ANOVAstandsforANalysis OfVAriance.• Usefulwhencomparingmorethan2means.• IfIhave2 meanstocompare,I justlookattheirdifferencetomeasurehowfaraparttheyare.

• SupposeI wantedtocomparethreemeans.IhavethemeanforgroupA,themeanforgroupB,andthemeanforgroupC.

Page 39: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Fteststatistic

• TheanalysisofvarianceF teststatisticis:

groupsy within variabilitgroupsbetween y variabilit

=F

• Thisissimilartothet-statisticwhenwewerecomparingjusttwomeans. 𝑡 = "̅d."̅g

kdg

hdlkd

g

hg

Page 40: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

RecallingAmbiguousProseExample9.2

Page 41: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ComprehensionExample

(Don’t follow along in your book or look ahead on the PowerPoint until after I read you the passage.) • Studentswerereadanambiguousprosepassageunderoneofthefollowingconditions:• Studentsweregivenapicturethatcouldhelptheminterpretthepassagebefore theyheardit.

• Studentsweregiventhepictureafter theyheardthepassage.

• Studentswerenot shownanypicturebeforeorafterhearingthepassage.

• Theywerethenaskedtoevaluatetheircomprehensionofthepassageona1to7scale.

Page 42: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ComprehensionExample• ThisexperimentisapartialreplicationdoneatHopeCollegeofastudy

donebyBransfordandJohnson(1972).• Studentswererandomlyassignedtooneofthe3groups.• Listentothepassageandseeifitmakessense.Wouldapicturehelp?Iftheballoonspopped,thesoundwouldn’tbeabletocarrysinceeverythingwouldbetoofarawayfromthecorrectfloor.Aclosedwindowwouldalsopreventthesoundfromcarrying,sincemostbuildingstendtobewellinsulated.Sincethewholeoperationdependsonasteadyflowofelectricity,abreakinthemiddleofthewirewouldalsocauseproblems.Ofcourse,thefellowcouldshout,butthehumanvoiceisnotloudenoughtocarrythatfar.Anadditionalproblemisthatastringcouldbreakontheinstrument.Thentherecouldbenoaccompanimenttothemessage.Itisclearthatthebestsituationwouldinvolvelessdistance.Thentherewouldbefewerpotentialproblems.Withfacetofacecontact,theleastnumberofthingscouldgowrong.

Page 43: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences
Page 44: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Hypotheses

• Null:Inthepopulationthereisnoassociationbetweenwhetherorwhenapicturewasshownandcomprehensionofthepassage

• Alternative:Inthepopulationthereisanassociationbetweenwhetherandwhenapicturewasshownandcomprehensionofthepassage

Page 45: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Hypotheses

• Null: All three of the long term mean comprehension scores are the same.

µnopicture =µpicturebefore =µpictureafter

• Alternative: At least one of the mean comprehension scores is different.

Page 46: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

RecallScore• Studentsratedtheircomprehension,andtheresearchersalsohadthestudentsrecallasmanyideasfromthepassageastheycould.Theywerethengradedonwhattheycouldrecallandtheresultsareshown.

Page 47: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ValidityConditions• Justaswiththesimulation-basedmethod,weareassumingwehaveindependentgroups.

• TwoextraconditionsmustbemettousetraditionalANOVA:• Normality:Ifsamplesizesaresmallwithineachgroup,datashouldn’tbeveryskewed.Ifitis,usesimulationapproach.(Samplesizesofatleast20isagoodguidelineformeans.)

• Equalvariation:Largeststandarddeviationshouldbenomorethantwicethevalueofthesmallest.

Page 48: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

ANOVAOutput• ThisisthekindofoutputyouwouldseeinmoststatisticspackageswhendoingANOVA.

• Thevariabilitybetweenthegroupsismeasuredbythemeansquaretreatment(40.02).

• Thevariabilitywithinthegroupsismeasuredbythemeansquareerror(3.16).

• TheFstatisticis40.02/3.16=12.67.

Page 49: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Conclusion• Sincewehaveasmallp-valuewehavestrongevidenceagainstthenullandcanconcludeatleastoneofthelong-runmeanrecallscoresisdifferent.

Page 50: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Reviewlist.1.MeaningofSD. 19.Randomsamplingandrandom2.Parametersandstatistics. assignment.3.Zstatisticforproportions. 20.TwoproportionCIsandtesting.4.Simulationandmeaningofpvalues. 21.IQRand5number summaries.5.SEforproportions. 22.CIs for 2means andtesting.6.Whatinfluencespvalues. 23.Paired data.7.CLTandvalidityconditionsfortests. 24.Placeboeffect,adhererbias,8.1-sidedand2-sidedtests. andnonresponsebias.9.Rejectthenullvs.acceptthealternative. 25.Predictionandcausation.10.Samplingandbias. 26.Multipletestingandpublicationbias11.Significancelevel. 27.Regression.12.TypeI,typeIIerrors,andpower. 28.Correlation.13.CIsforaproportion. 29.Calculate&interpreta&b.14.CIsforamean. 30.Goodnessoffitforregression.15.Marginoferror. 31.Commonregressionproblems16.Practicalsignificance. (causation,extrapolation,curvature,heterskedasticity).17.Confounding. 32.ANOVAandF-test.18.Observationalstudiesandexperiments.

Page 51: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

1.Whatdoesthecorrelationof0.82imply?a.82%ofthevariationinweightisexplainedbyheight.b.Thetypicalvariationinpeople'sheightsis82%aslargeasthetypicalvariationintheirweights.c.Thereisstrongassociationbetweenheightandweightinthissample.d.Foreveryinchofincreaseinone'sheight,wewouldpredicta0.82lb.increaseinweight.e.Ifapersonweighs100pounds,thenwetypicallywouldexpectthepersontobeabout82inchestall.

Page 52: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

1.Whatdoesthecorrelationof0.82imply?a.82%ofthevariationinweightisexplainedbyheight.b.Thetypicalvariationinpeople'sheightsis82%aslargeasthetypicalvariationintheirweights.c.Thereisstrongassociationbetweenheightandweightinthissample.d.Foreveryinchofincreaseinone'sheight,wewouldpredicta0.82lb.increaseinweight.e.Ifapersonweighs100pounds,thenwetypicallywouldexpectthepersontobeabout82inchestall.

Page 53: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

2.Whatistheestimatedslope,inlbs/inch,oftheregressionlineforpredictingweightfromheight?a.6.56.b.7.12.c.8.04.d.9.92.e.10.2.f.11.4.

Page 54: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

2.Whatistheestimatedslope,inlbs/inch,oftheregressionlineforpredictingweightfromheight?a.6.56.b.7.12.c.8.04.d.9.92.e.10.2.f.11.4.

rsy/sx =.82x 40/5=6.56.

Page 55: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

3.Howmuchwouldapredictionusingthisregressionlinetypicallybeoffby?a.12.7lbs. b.13.5lbs. c.14.4lbs. d.20.2lbs.e.22.9lbs.

Page 56: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

3.Howmuchwouldapredictionusingthisregressionlinetypicallybeoffby?a.12.7lbs. b.13.5lbs. c.14.4lbs. d.20.2lbs.e.22.9lbs.

√(1-r2)sy =√(1-.822)x40=22.9.

Page 57: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

4.Ifweweretorandomlytakeoneadultfromthissample,howmuchwouldhis/herheighttypicallydifferfrom65by?a.0.05in.b.0.1in.c.0.5in.d.1.0in.e.2.5in.f.5.0in.

Page 58: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

exampleproblems.

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

4.Ifweweretorandomlytakeoneadultfromthissample,howmuchwouldhis/herheighttypicallydifferfrom65by?a.0.05in.b.0.1in.c.0.5in.d.1.0in.e.2.5in. f.5.0in.

Page 59: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,themedianheightis64.5inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

5.Whyshouldn'tonetrustthisregressionlinetopredicttheweightofsomeonewhois25inchestall?a.Thesamplesizeisinsufficientlylarge.b.ThesampleSDofweightistoosmall.c.Thevalueof25inchesistoofaroutsidetherangeofmostobservations.d.ThecorrelationoftheANOVAisat-testconfidenceintervalwithstatisticalsignificance.e.Thedatacomefromanobservationalstudy,sotheremaybeconfoundingfactors.f.Theheightvaluesareheavilyrightskewed,sothepredictionerrorsarelarge.

Page 60: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,themedianheightis64.5inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.

5.Whyshouldn'tonetrustthisregressionlinetopredicttheweightofsomeonewhois25inchestall?a.Thesamplesizeisinsufficientlylarge.b.ThesampleSDofweightistoosmall.c.Thevalueof25inchesistoofaroutsidetherangeofmostobservations.d.ThecorrelationoftheANOVAisat-testconfidenceintervalwithstatisticalsignificance.e.Thedatacomefromanobservationalstudy,sotheremaybeconfoundingfactors.f.Theheightvaluesareheavilyrightskewed,sothepredictionerrorsarelarge.

Page 61: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,themedianheightis64.5inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.Theestimatedslopeinpredictingweightfromheightis6.56lbs/in.

6.Howshouldoneinterprettheestimatedslopeof6.56?a.Eachextrainchyougrowcausesyoutoincreaseyourweightby6.56lbs onaverage.b.Eachextralb.youweighcausesyoutogrow6.56inches.c.TheamountofweightAmericansaverageis6.56standarderrorsabovethemean.d.TheZ-scorecorrespondingtothecorrelationbetweenheightandweightis6.56.e.Foreachextrainchtalleryouare,yourpredictedweightincreasesby6.56lbs.f.Theproportionofvarianceinweightexplainedbytheregressionequationis6.56%.

Page 62: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposethatamongasampleof100adultsinagiventown,thecorrelationbetweenheight(inches)andweight(lbs.)is0.82,andthemeanheightis65inches,themedianheightis64.5inches,thesd ofheightis5inches,themeanweightis160lbs.,andthesd ofweightis40lbs.Theestimatedslopeinpredictingweightfromheightis6.56lbs/in.

6.Howshouldoneinterprettheestimatedslopeof6.56?a.Eachextrainchyougrowcausesyoutoincreaseyourweightby6.56lbs onaverage.b.Eachextralb.youweighcausesyoutogrow6.56inches.c.TheamountofweightAmericansaverageis6.56standarderrorsabovethemean.d.TheZ-scorecorrespondingtothecorrelationbetweenheightandweightis6.56.e.Foreachextrainchtalleryouare,yourpredictedweightincreasesby6.56lbs.f.Theproportionofvarianceinweightexplainedbytheregressionequationis6.56%.

Page 63: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

7.Whatisthepooledsamplepercentagewithacne,inbothgroupscombined?a.16.1%.b.17.2%.c.18.4%.d.19.7%.e.21.2%.f.23.0%.

Page 64: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

7.Whatisthepooledsamplepercentagewithacne,inbothgroupscombined?a.16.1%.b.17.2%.c.18.4%.d.19.7%. e.21.2%.f.23.0%.

118/600=19.7%.

Page 65: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

8.Usingthispooledsamplepercentage,underthenullhypothesisthatthetwogroupshavethesameacnerate,whatisthestandarderrorforthedifferencebetweenthetwopercentages?a.1.42%. b.1.88%. c.2.02%. d.2.99%.e.3.08%. f.3.44%.

Page 66: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

8.Usingthispooledsamplepercentage,underthenullhypothesisthatthetwogroupshavethesameacnerate,whatisthestandarderrorforthedifferencebetweenthetwopercentages?a.1.42%. b.1.88%. c.2.02%. d.2.99%.e.3.08%. f.3.44%.

√(.197*(1-.197)/400+.197*(1-.197)/200)=.0344.

Page 67: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

9.WhatistheZstatisticforthedifferencebetweenthetwogrouppercentages?a.1.02. b.1.23. c.1.55. d.1.88. e.2.03. f.2.43.

Page 68: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

9.WhatistheZstatisticforthedifferencebetweenthetwogrouppercentages?a.1.02. b.1.23. c.1.55. d.1.88. e.2.03. f.2.43.

(88/400– 30/200)÷ 0.0344=2.03.

Page 69: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

10.Usingtheunpooled standarderrorforthedifferencebetweenthetwopercentages,finda95%confidenceintervalforthepercentagewithacneamongthoseage21-30minusthepercentageofacneamongthoseage31-40.a.7%+/- 5.02%.b.7%+/- 5.54%.c.7%+/- 5.92%.d.7%+/- 6.03%.e.7%+/- 6.41%.

Page 70: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

10.Usingtheunpooled standarderrorforthedifferencebetweenthetwopercentages,finda95%confidenceintervalforthepercentagewithacneamongthoseage21-30minusthepercentageofacneamongthoseage31-40.a.7%+/- 5.02%.b.7%+/- 5.54%.c.7%+/- 5.92%.d.7%+/- 6.03%.e.7%+/- 6.41%.

88/400– 30/200+/- 1.96*√(88/400*(1-88/400)/400+30/200*(1-30/200)/200)=7%+/- 6.41%.

Page 71: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

11.Whatcanweconcludefromthe95%CIintheproblemabove?a.Thereisnostatisticallysignificantdifferenceinthepercentagewithacneamongthoseage21-30andthoseage31-40.b.Thereisstatisticallysignificantcorrelationbetweenthesamplesizeandtheeffectsizefortheconfoundingfactorsofarandomizedcontrolledexperiment.c.Ofthosewithacne,thereisnostatisticallysignificantdifferencebetweenthepercentagewhoareage21-30andthepercentagewhoare31-40.d.Astatisticallysignificantlyhigherpercentageofpeopleage21-30haveacnethanthoseage31-40.e.Thesamplesizesaretoosmallforaconclusiontobevalidhere.

Page 72: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Supposearesearcherstudyingacne,oroilyskin,takesasimplerandomsampleof400Americansage21-30andcallsthemgroupA,andasimplerandomsampleof200Americansage31-40andcallsthemgroupB.IngroupA,88peoplehaveacne,andingroupB,30peoplehaveacne.

11.Whatcanweconcludefromthe95%CIintheproblemabove?a.Thereisnostatisticallysignificantdifferenceinthepercentagewithacneamongthoseage21-30andthoseage31-40.b.Thereisstatisticallysignificantcorrelationbetweenthesamplesizeandtheeffectsizefortheconfoundingfactorsofarandomizedcontrolledexperiment.c.Ofthosewithacne,thereisnostatisticallysignificantdifferencebetweenthepercentagewhoareage21-30andthepercentagewhoare31-40.d.Astatisticallysignificantlyhigherpercentageofpeopleage21-30haveacnethanthoseage31-40.e.Thesamplesizesaretoosmallforaconclusiontobevalidhere.

Page 73: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Forthefollowingtwoproblems,supposearesearcherstudiesatreatmentforacertaincondition.Theresearcherdivides230subjectsrandomlyintothreegroups.GroupAreceivesaplacebo.GroupBreceivesthetreatmentatalowdose,andgroupCreceivesthetreatmentatahighdose.ThetablebelowshowstheoutputfromanANOVAF-testonthemeanlevelsoftheconditioninthe3groups.

AnalysisofVarianceTable

Response:outcomeDf SumSq MeanSq Fvalue Pr(>F)

group 20.467 0.233 0.221 0.803Residuals 22728.5 1.06

12.Inthetable,whatnumberisameasureofthevariabilitybetweengroups?a.2.b.0.233c.0.221d.0.803e.28.5 f.1.06.

Page 74: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Forthefollowingtwoproblems,supposearesearcherstudiesatreatmentforacertaincondition.Theresearcherdivides230subjectsrandomlyintothreegroups.GroupAreceivesaplacebo.GroupBreceivesthetreatmentatalowdose,andgroupCreceivesthetreatmentatahighdose.ThetablebelowshowstheoutputfromanANOVAF-testonthemeanlevelsoftheconditioninthe3groups.

AnalysisofVarianceTable

Response:outcomeDf SumSq MeanSq Fvalue Pr(>F)

group 20.467 0.233 0.221 0.803Residuals 22728.5 1.06

12.Inthetable,whatnumberisameasureofthevariabilitybetweengroups?a.2.b.0.233 c.0.221d.0.803e.28.5 f.1.06.

Page 75: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Forthefollowingtwoproblems,supposearesearcherstudiesatreatmentforacertaincondition.Theresearcherdivides230subjectsrandomlyintothreegroups.GroupAreceivesaplacebo.GroupBreceivesthetreatmentatalowdose,andgroupCreceivesthetreatmentatahighdose.ThetablebelowshowstheoutputfromanANOVAF-testonthemeanlevelsoftheconditioninthe3groups.

AnalysisofVarianceTableResponse:outcome

Df SumSq MeanSq Fvalue Pr(>F)group 20.467 0.233 0.221 0.803Residuals 22728.5 1.06

13.WhatcanweconcludefromtheresultoftheFtest?a.Thetreatmenthasastatisticallysignificanteffect,thoughwecannotbesurewhichdoseisattributabletothissignificanteffectorinwhichdirectiontheeffectgoes.b.Thetreatmentseemstohavesignificantlygreatereffectatlargedosesthanatsmalldoses.c.Thetreatmentdoesnotseemtohaveastatisticallysignificanteffect.d.Wefailtorejectthenullhypothesisthatthetreatmenthasasignificanteffectonthecondition.e.Thecorrelationbetweendoseandoutcomeappearstobestatisticallysignificant.

Page 76: Stat 13, Intro. to Statistical Methods for the Life and Health …frederic/13/sum17/day09.pdf · 2017-07-31 · Stat 13, Intro. to Statistical Methods for the Life and Health Sciences

Forthefollowingtwoproblems,supposearesearcherstudiesatreatmentforacertaincondition.Theresearcherdivides230subjectsrandomlyintothreegroups.GroupAreceivesaplacebo.GroupBreceivesthetreatmentatalowdose,andgroupCreceivesthetreatmentatahighdose.ThetablebelowshowstheoutputfromanANOVAF-testonthemeanlevelsoftheconditioninthe3groups.

AnalysisofVarianceTableResponse:outcome

Df SumSq MeanSq Fvalue Pr(>F)group 20.467 0.233 0.221 0.803Residuals 22728.5 1.06

13.WhatcanweconcludefromtheresultoftheFtest?a.Thetreatmenthasastatisticallysignificanteffect,thoughwecannotbesurewhichdoseisattributabletothissignificanteffectorinwhichdirectiontheeffectgoes.b.Thetreatmentseemstohavesignificantlygreatereffectatlargedosesthanatsmalldoses.c.Thetreatmentdoesnotseemtohaveastatisticallysignificanteffect.d.Wefailtorejectthenullhypothesisthatthetreatmenthasasignificanteffectonthecondition.e.Thecorrelationbetweendoseandoutcomeappearstobestatisticallysignificant.