You are on page 1of 11

Pagei

Analyzingneuraltimeseriesdata:
Theoryandpractice

MikeXCohen

MITPress,early2014


Pageii

Contents

Section1:Introductions
Chapter1:Thepurposeofthisbook,whoshouldreadit,andhowtouseit
1.1Whatiscognitiveelectrophysiology?
1.2Whatisthepurposeofthisbook?
1.3Whyshouldntyouuse<insertnameofM/EEGsoftwareanalysispackage>?
1.4Whyprogramanalyses,andwhyinMatlab?
1.5Howbesttolearnfromandusethebook
1.6Sampledataandonlinecode
1.7Terminologyusedinthisbook
1.8Exercises
1.9IseverythingthereistoknowaboutEEGanalysesinthisbook?
1.10Whoshouldreadthisbook?
1.11Isthisbookdifficult?
1.12Questions?

Chapter2:Advantagesandlimitationsoftimeandtimefrequencydomainanalyses
2.1WhyEEG?
2.2WhynotEEG?
2.3InterpretingvoltagevaluesfromtheEEGsignal
2.4Advantagesofeventrelatedpotentials(ERPs)
2.5LimitationsofERPs
2.6Advantagesoftimefrequencybasedapproaches
2.7Limitationsoftimefrequencybasedapproaches
2.8Temporalresolution,precision,andaccuracyofEEG
2.9Spatialresolution,precision,andaccuracyofEEG
2.10Topographicallocalizationvs.brainlocalization
2.11EEGorMEG?
2.12CostsofEEGresearch

Chapter3:Interpretingandaskingquestionsabouttimefrequencyresults
3.1EEGtimefrequency:Thebasics
3.2Waystoviewtimefrequencyresults
3.3tfviewerxanderpviewerx
3.4Howtoviewandinterprettimefrequencyresults
3.5Thingstobesuspiciousofwhenviewingtimefrequencyresults
3.6Doresultsintimefrequencyplotsmeanthattherewereneuraloscillations?

Chapter4:IntroductiontoMatlabprogramming
4.1Writecleanandefficientcode
4.2Usemeaningfulfileandvariablenames
4.3Makeregularbackupsofyourcode,andkeeporiginalcopiesofmodifiedcode
4.4Initializevariables
Pageiii

4.5Help!
4.6Bepatientandembracethelearningexperience
4.7Exercises

Chapter5:IntroductiontothephysiologicalbasesofEEG
5.1BiophysicaleventsthataremeasurablewithEEG
5.2Neurobiologicalmechanismsofoscillations
5.3Phaselocked,timelocked,taskrelated
5.4NeurophysiologicalmechanismsofERPs
5.5Areelectricalfieldscausallyinvolvedincognition?
5.6Whatifelectricalfieldsarenotcausallyinvolvedincognition?

Chapter6:PracticalitiesofEEGmeasurementandexperimentaldesign
6.1Designingexperiments:discuss,pilot,discuss,pilot
6.2Eventmarkers
6.3Intraandintertrialtiming
6.4Howmanytrialsyouwillneed
6.5Howmanyelectrodesyouwillneed
6.6Whichsamplingratetousewhenrecordingdata
6.7Otheroptionalequipmenttoconsider

Section2:Preprocessingandtimedomainanalyses
Chapter7:Preprocessingstepsnecessaryandusefulforadvanceddataanalysis
7.1Whatispreprocessing?
7.2Thebalancebetweensignalandnoise
7.3Creatingepochs
7.4Matchingtrialcountacrossconditions
7.5Filtering
7.6Trialrejection
7.7Spatialfiltering
7.8Referencing
7.9Interpolatingbadelectrodes
7.10Startwithcleandata

Chapter8:EEGartifacts:theirdetection,influence,andremoval
8.1Removingdatabasedonindependentcomponentsanalysis
8.2Removingtrialsbecauseofblinks
8.3Removingtrialsbecauseofoculomotoractivity
8.4RemovingtrialsbasedonEMGinEEGchannels
8.5RemovingtrialsbasedonresponsehandEMG
8.6Removingtrialsbasedontaskperformance
8.7Trainsubjectstominimizeartifacts
8.8Minimizeartifactsduringdatacollection

Pageiv

Chapter9:OverviewoftimedomainEEGanalyses
9.1Eventrelatedpotentials
9.2FilteringERPs
9.3Butterflyplotsandglobalfieldpower/topographicalvarianceplots
9.4Theflickereffect
9.5Topographicalmaps
9.6Microstates
9.7ERPimages
9.8Exercises

Section3:Frequencyandtimefrequencydomainsanalyses
Chapter10:Thedotproductandconvolution
10.1Dotproduct
10.2Convolution
10.3Howdoesconvolutionwork?
10.4Convolutionvs.crosscovariance
10.5ThepurposeofconvolutionforEEGdataanalyses
10.6Exercises

Chapter11:ThediscretetimeFouriertransform,theFFT,andtheconvolutiontheorem
11.1Makingwaves
11.2FindingwavesinEEGdatawiththeFouriertransform
11.3ThediscretetimeFouriertransform
11.4VisualizingtheresultsofaFouriertransform
11.5Complexresultsandnegativefrequencies
11.6InverseFouriertransform
11.7TheFastFourierTransform
11.8StationarityandtheFouriertransform
11.9Extractingmoreorfewerfrequenciesthandatapoints
11.10Theconvolutiontheorem
11.11PerformingFFTbasedconvolutioninMatlab
11.12Exercises

Chapter12:Morletwaveletsandwaveletconvolution
12.1Whywavelets?
12.2Howtomakewavelets
12.3Waveletconvolutionasabandpassfilter
12.4Limitationsofwaveletconvolutionasdiscussedthusfar
12.5Exercises

Chapter13:Complexwaveletsandextractingpowerandphase
13.1Thewaveletcomplex
13.2Imaginingtheimaginary
13.3Rectangularandpolarnotation,andthecomplexplane
Pagev

13.4Eulersformula
13.5Eulersformulaandtheresultofcomplexwaveletconvolution
13.6Fromtimepointtotimeseries
13.7Parametersofwaveletsandrecommendedsettings
13.8Determiningthefrequencysmoothingofwavelets
13.9TipsforwritingefficientconvolutioncodeinMatlab
13.10DescribingthisanalysisinyourMethodssection
13.11Exercises

Chapter14:BandpassfilteringandtheHilberttransform
14.1Hilberttransform
14.2FilteringdatabeforeapplyingtheHilberttransform
14.3Finitevs.infiniteimpulseresponsefilters
14.4Bandpass,bandstop,highpass,lowpass
14.5Constructingafilter
14.6Checkyourfilters
14.7Applyingthefiltertodata
14.8Butterworth(IIR)filter
14.9Filteringeachtrialvs.filteringconcatenatedtrials
14.10Multiplefrequencies
14.11Aworldoffilters
14.12DescribingthisanalysisinyourMethodssection
14.13Exercises

Chapter15:ShorttimeFFT
15.1HowtheshorttimeFFTworks
15.2Taperthetimeseries
15.3Timesegmentlengthsandoverlap
15.4Powerandphase
15.5DescribingthisanalysisinyourMethodssection
15.6Exercises

Chapter16:Multitaper
16.1Howthemultitapermethodworks
16.2Thetapers
16.3Whenyoushouldandshouldnotusemultitapers
16.4TheMultitaperframeworkandadvancedtopics
16.5DescribingthisanalysisinyourMethodssection
16.6Exercises

Chapter17:Lesscommonlyusedtimefrequencydecompositionmethods
17.1Autoregressivemodeling
17.2HilbertHuang
17.3Matchingpursuit
Pagevi

17.4Pepisode
17.6Stransform

Chapter18:Timefrequencypower,andbaselinecorrections
18.11/fpowerscaling
18.2Thesolutionto1/fpowerintaskdesigns
18.3Decibelconversion
18.4Percentchangeandbaselinedivision
18.5Ztransform
18.6Notalltransformsareequal
18.7Othertransforms
18.8Meanvs.median
18.9Singletrialbaselinenormalization
18.10Thechoiceofbaselinetimewindow
18.11Disadvantagesofbaselinenormalizedpower
18.12Signaltonoiseestimates
18.13Numberoftrialsandpowerestimates
18.14Downsamplingresultsafteranalyses
18.15DescribingthisanalysisinyourMethodssection
18.16Exercises

Chapter19:Intertrialphaseclustering
19.1Whyphasevaluescannotbeaveraged
19.2Intertrialphaseclustering
19.3Strengthinnumbers
19.4UsingITPCwhentherearefewtrialsorconditiondifferencesintrialcount
19.5EffectsoftemporaljitteronITPCandpower
19.6ITPCandpower
19.7WeightedITPC
19.8Multimodalphasedistributions
19.9Spikefieldcoherence
19.10DescribingthisanalysisinyourMethodssection
19.11Exercises

Chapter20:Differencesamongtotal,phaselocked,andnonphaselockedpower,andphase
clustering
20.1Totalpower
20.2Nonphaselockedpower
20.3Phaselockedpower
20.4Intertrialphaseclustering
20.5Whentousewhatapproach
20.6Exercises

Chapter21:Interpretationsandlimitationsoftimefrequencypowerandphaseanalyses
Pagevii

21.1Terminology
21.2Whentousewhattimefrequencydecompositionmethod
21.3Interpretingtimefrequencypower
21.4Interpretingtimefrequencyintertrialphaseclustering
21.5Limitationsoftimefrequencypowerandintertrialphaseclustering
21.6Dotimefrequencyanalysesrevealneuraloscillations?

Section4:Spatialfilters
Chapter22:SurfaceLaplacian
22.1WhatisthesurfaceLaplacian?
22.2AlgorithmsforcomputingthesurfaceLaplacianforEEGdata
22.3SurfaceLaplacianfortopographicallocalization
22.4SurfaceLaplacianforconnectivityanalyses
22.5SurfaceLaplacianforcleaningtopographicalnoise
22.6DescribingthisanalysisinyourMethodssection
22.7Exercises

Chapter23:Principalcomponentsanalysis
23.1PurposeandinterpretationsofPCA
23.2HowPCAiscomputed
23.3Distinguishingsignificantfromnonsignificantcomponents
23.4RotatingPCAsolutions
23.5TimeresolvedPCA
23.6PCAwithtimefrequencyinformation
23.7PCAacrossconditions
23.8Independentcomponentsanalysis(ICA)
23.9Describingthismethodinyourmethodssection
23.10Exercises

Chapter24:Basicsofsingledipoleanddistributedsourceimaging
24.1Theforwardsolution
24.2Theinverseproblem
24.3Dipolefitting
24.4Nonadaptivedistributedsourceimagingmethods

Section5:Connectivity
Chapter25:Introductiontothevariousconnectivityanalyses
25.1Whyonlytwosites(bivariateconnectivity)?
25.2Importantconceptsrelatedtobivariateconnectivity
25.3Whichmeasureofconnectivityshouldbeused?
25.4Phasebasedconnectivity
25.5Powerbasedconnectivity
25.6Grangerprediction
25.7Mutualinformation
Pageviii

25.8Crossfrequencycoupling
25.9Graphtheory
25.10Potentialconfoundofvolumeconduction

Chapter26:Phasebasedconnectivity
26.1Terminology
26.2Intersitephaseclustering(ISPC)overtime
26.3ISPCtrials
26.4ISPCandthenumberoftrials
26.5RelationbetweenISPCandpower
26.6WeightedISPCtrials
26.7Spectralcoherence(a.k.a.magnitudesquaredcoherence)
26.8Imaginarycoherence,phaselagindex,andweightedphaselagindex
26.9Whichmeasureofphaseconnectivityshouldyouuse?
26.10Testingthemeanphaseangle
26.11DescribingtheseanalysesinyourMethodssection
26.12Exercises

Chapter27:Powerbasedconnectivity
27.1Spearmanvs.Pearsoncoefficientforpowercorrelations
27.2Powercorrelationsovertime
27.3Powercorrelationsovertrials
27.4Partialcorrelations
27.5Matlabprogrammingtips
27.6DescribingthisanalysisinyourMethodssection
27.7Exercises

Chapter28:Grangerprediction
28.1Univariateautoregression
28.2Bivariateautoregression
28.3Autoregressionerrorsanderrorvariances
28.4Grangerpredictionovertime
28.5Modelorder
28.6FrequencydomainGrangerprediction
28.7Timeseriescovariancestationarity
28.8BaselinenormalizationofGrangerpredictionresults
28.9Statistics
28.10AdditionalapplicationsofGrangerprediction
28.11Exercises

Chapter29:Mutualinformation
29.1Entropy
29.2Howmanyhistogrambinstouse
29.3Enjoytheentropy
Pageix

29.4Jointentropy
29.5Mutualinformation
29.6Mutualinformationandamountofdata
29.7Mutualinformationwithnoisydata
29.8Mutualinformationovertimeorovertrials
29.9Mutualinformationonrealdata
29.10Mutualinformationonfrequencybandspecificdata
29.11Laggedmutualinformation
29.12Statistics
29.13Moreinformation
29.14DescribingthisanalysisinyourMethodssection
29.15Exercises

Chapter30:Crossfrequencycoupling
30.1Visualinspectionofcrossfrequencycoupling
30.2Powerpowercorrelations
30.3Aprioriphaseamplitudecoupling
30.4Separatingtaskrelatedphaseandpowercoactivationsfromphaseamplitudecoupling
30.5Mixedapriori/exploratoryphaseamplitudecoupling
30.6Exploratoryphaseamplitudecoupling
30.7Notesaboutphaseamplitudecoupling
30.8Phasephasecoupling
30.9Othermethodsforquantifyingcrossfrequencycoupling
30.10Crossfrequencycouplingovertimeorovertrials
30.11DescribingthisanalysisinyourMethodssection
30.12Exercises

Chapter31:Graphtheory
31.1Networksasmatricesandgraphs
31.2Thresholdingconnectivitymatrices
31.3Connectivitydegree
31.3Clusteringcoefficient
31.4Pathlength
31.5Smallworldnetworks
31.6Statistics
31.7Howtodescribetheseanalysesinyourpaper
31.8Exercises

Section6:Statisticalanalyses
Chapter32:Advantagesandlimitationsofdifferentstatisticalprocedures
32.1Arestatisticsnecessary?
32.2Atwhatlevelshouldstatisticsbeperformed?
32.3Whatpvalue(orothersignificancethreshold)shouldbeused,andshouldmultiple
comparisonscorrectionsbeapplied?
Pagex

32.4Arepvaluestheonlystatisticalmetric?
32.5Statisticalsignificancevs.practicalsignificance
32.6TypeIandTypeIIerrors
32.7Whatkindsofstatistics(e.g.,parametric,nonparametric,Bayesian)shouldbeapplied?
32.8Howtocombinedataacrosssubjects

Chapter33:Nonparametricpermutationtesting
33.1Advantagesofnonparametricpermutationtesting
33.2Creatinganullhypothesisdistribution
33.3Howmanyiterationsarenecessaryforthenullhypothesisdistribution?
33.4Determiningstatisticalsignificance
33.5Multiplecomparisonsandtheircorrections
33.6Correctionformultiplecomparisonsatthepixellevel
33.7Correctionsformultiplecomparisonsattheclusterlevel
33.8Falsediscoveryrate(FDR)formultiplecomparisonscorrection
33.9Whatshouldbepermuted?
33.10Nonparametricpermutationtestingbeyondsimplebivariatecases
33.11DescribingthisanalysisinyourMethodssection

Chapter34:Withinsubjectstatisticalanalyses
34.1Changesintaskrelatedpowercomparedtobaseline
34.2Discreteconditiondifferencesinpower
34.3Continuousrelationshipwithpower:singletrialcorrelations
34.4Continuousrelationshipswithpower:singletrialmultipleregression
34.5Determiningstatisticalsignificanceofphasebaseddata(ITPCandISPC)
34.6Testingpreferredphaseangleacrossconditions
34.7Testingthestatisticalsignificanceofcorrelationcoefficients

Chapter35:Grouplevelanalysesandappropriatedataanalysisstrategies
35.1Avoidcircularinferences
35.2Grouplevelanalysisstrategy1:Testeachpixelandapplyamapwisethreshold
35.3Grouplevelanalysisstrategy2a:Timefrequencywindowsforhypothesisdrivenanalyses
35.4Grouplevelanalysisstrategy2b:Subjectspecifictimefrequencywindowsforhypothesis
drivenanalyses
35.5Determininghowmanysubjectsyouneedforgrouplevelanalyses

Chapter36:Recommendationsforreportingresultsinfigures,tables,andtext
36.1Recommendation1:Onefigure,oneidea
36.2Recommendation2:Showdata
36.3Recommendation3:Highlightsignificanteffectsinsteadofremovingnonsignificant
effects
36.4Recommendation4:Showspecificity(orlackthereof)infrequency,time,andspace
36.5Recommendation5:Usecolor
36.6Recommendation6:Useinformativefigurelabelsandcaptions
Pagexi

36.7Recommendation7:Avoidshowingrepresentativedata
36.8Achecklistformakingfigures
36.9Tables
36.10ReportingresultsintheResultssection

Section7:Conclusionsandfuturedirections
Chapter37:Recurringthemesinthisbook,andsomepersonaladvice
37.1Theme:Myriadpossibleanalyses
37.2Advice:Avoidtheparalysisofanalysis
37.3Theme:Youdonthavetoprogramyourownanalyses,butyoushouldknowhowanalyses
work
37.4Advice:Ifitfeelswrong,itprobablyis
37.5Advice:Whenindoubt,plotitout
37.6Advice:Knowthesethreeformulaelikethebackofyourhand
37.7Theme:Connectivityovertrialsorovertime
37.8Theme:Mostanalysisparametersintroducebias
37.9Theme:WriteaclearMethodssectionsootherscanreplicateyouranalyses
37.10Theme:Usedescriptiveandappropriateanalysisterms
37.11Advice:Interpretnullresultscautiously
37.12Advice:Trysimulations,butalsotrustrealdata
37.13Advice:Trustreplications
37.14Theme:Analysesarenotrightorwrong;theyareappropriateorinappropriate
37.15Advice:Hypothesistestingisgood/bad,andsoisdatadrivenexploration
37.16Advice:Findsomethingthatdrivesyou,andstudyit
37.17Cognitiveelectrophysiology:Theartoffindinganthillsonmountains

Chapter38:Thefutureofcognitiveelectrophysiology
38.1Developmentsinanalysismethods
38.2DevelopmentsinunderstandingtheneurophysiologyofEEG
38.3Developmentsinexperimentdesign
38.4Developmentsinmeasurementtechnology
38.5Theroleofthebodyinbrainfunction
38.6Determiningcausality
38.7InferringcognitivestatesfromEEGsignatures(inverseinference)
38.8Tablesofactivation
38.9Diseasediagnosisandpredictingtreatmentcourseandsuccess
38.10Clinicalrelevanceisnotnecessaryfortheadvancementofscience
38.11Replications
38.12Tripleblindreviewforscientificpublications
38.13?

You might also like