Modelling bovine trypanosomosis spatial distribution by GIS in an agro pastoral zone of Burkina Faso

pages

English

Documents scolaires

Écrit par
Jean - Franc¸Ois Michela

Publié par
profil-ziem-2012

Lire

Le téléchargement nécessite un accès à la bibliothèque YouScribe Tout savoir sur nos offres

pages

English

Ebook

Lire

Le téléchargement nécessite un accès à la bibliothèque YouScribe Tout savoir sur nos offres

Publié par

profil-ziem-2012

Nombre de lectures

Langue

English

Niveau: Secondaire, Lycée, Terminale
UN CO RR EC TE D PR OO 3 Modelling bovine trypanosomosis spatial 4 distribution by GIS in an agro-pastoral 5 zone of Burkina Faso 6 Jean-Franc¸ois Michela,*, Stephane Drayb, Stephane de La Rocquea, 7 Marc Desquesnesa, Philippe Solanoc, Gerard De Wispelaered, 8 Dominique Cuisanceee 9 aCIRDES/CIRAD-EMVT, Bobo Dioulasso, Burkina Faso, France 10 bUniversite Claude Bernard Lyon I, Villeurbanne, France 11 cIRD/IPR, Bouake, Cote d'Ivoire, France 12 dCIRAD-EMVT, Maison de la teledetection, Montpellier, France 13 eCIRAD-EMVT, Campus International de Baillarguet, Montpellier, France 14 15 Abstract 16 17 Modelling of the spatial distribution of bovine trypanosomosis prevalence in Sideradougou district 18 Burkina Faso was performed by using a combination of spatial and statistical analysis. Based on a 19 comprehensive and geographically representative census of herds and farms in the area, more than 20 2000 cattle were randomly chosen and their blood sampled during field survey. Data on livestock 21 farming practices were recorded for each farm. All data were mapped within a GIS to generate new 22 information on spatial constraints in the area. 23 Surveys results were analysed and serological prevalence data were modelled using logistic 24 regression. The model allowed identification and quantification of risk factors. In a second step the 25 statistical model was used predictively on the entire farm population in the area.

modelling bovine

modelling disease spatial

trypanosomosis

eralized linear model

head

practised during

systems practised

function used

Voir

Publié par

profil-ziem-2012

Nombre de lectures

Langue

English

3456789011121314151617181910212223242526272829203132333435312PreventiveVeterinaryMedicine1728(2002)1–14ModellingbovinetrypanosomosisspatialdistributionbyGISinanagro-pastoralzoneofBurkinaFasoJean-Franc¸oisMichela,*,St´ephaneDrayb,St´ephanedeLaRocquea,MarcDesquesnesa,PhilippeSolanoc,G´erardDeWispelaered,DominiqueCuisanceeeaCIbRUDnEivS/erCsIitR´eADCl-aEuMdeVTB,eBrnoabrodDLiyoounlaIs,sVoi,llBeuurrkbiannaneF,asFor,anFcreancecIRD/IPR,Bouake,Cˆoted’Ivoire,FrancedeCIRACDI-REAMDV-TE,MCVaTm,pMusaiIsnotnerdneatliaontae´ll´eddeetBeactiilloanr,guMeot,ntMpeolnlitepre,llFirear,ncFeranceAbstractModellingofthespatialdistributionofbovinetrypanosomosisprevalenceinSideradougoudistrictBurkinaFasowasperformedbyusingacombinationofspatialandstatisticalanalysis.Basedonacomprehensiveandgeographicallyrepresentativecensusofherdsandfarmsinthearea,morethan2000cattlewererandomlychosenandtheirbloodsampledduringﬁeldsurvey.Dataonlivestockfarmingpracticeswererecordedforeachfarm.AlldataweremappedwithinaGIStogeneratenewinformationonspatialconstraintsinthearea.Surveysresultswereanalysedandserologicalprevalencedataweremodelledusinglogisticregression.Themodelallowedidentiﬁcationandquantiﬁcationofriskfactors.Inasecondstepthestatisticalmodelwasusedpredictivelyontheentirefarmpopulationinthearea.Thismethodwassuccessfulinpredictingtheserologicalprevalenceforeachindividualherdinthesample,fromtheirlivestockmanagementpatternsandspatiallocation.PredictedprevalenceswererepresentedwithintheGIS,takingdailymovementsofanimalsintoaccount.Spatialdistributionofprevalencewouldillustratespeciﬁclocationsatriskfromanepidemiologicalviewpoint.Itgivesevidencethatthehydrologicalnetworkandlandoccupationpatternsinthesavanna-typecountrysideareplayinganimportantpartwhenstructuringaso-called‘‘trypanosomosisspace’’.#2002PublishedbyElsevierScienceB.V.Keywords:GIS;Spatialmodelling;Logisticregression;Trypanosomosis;Epidemiology*Correspondingauthor.E-mailaddress:jefmichel@wanadoo.fr(J.-F.Michel).0167-5877/02/$–seefrontmatter#2002PublishedbyElsevierScienceB.V.PII:S0167-5877(02)00120-4

63738393041424344454647484940515253545556575859506162636465666768696071727372J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–141.IntroductionAnimaltrypanosomosesareoneofthemainpathologicalconstraintsonthedevelopmentofanimalproductioninsub-SaharanAfrica(HurseyandSlingenbergh,1995),andcauseannuallossesestimatedatUS$1billion(DeHaanandBekure,1991).Tsetseﬂiesarethemainvectors.Theriskoftransmissionisprimarilylinkedtotheintensityoftheencountersbetweenvectorsandhosts,anddependsonthespatialandtemporalinterfacesbetweentheprotagonistsinthepathogensystem(host–vector–parasite)(Laveissi`ereetal.,1986;DeLaRocqueetal.,1999).High-riskareashavebeenidentiﬁedonthisbasisinanagro-pastoralzoneofsouthernBurkinaFaso,takingenvironmentalandsocio-economicfactorsintoaccount.Theavailabledataweregeoreferenced,includedintoageographicinformationsystem(GIS),andhigh-riskareaswereidentiﬁedbyspatialmodelling(DeLaRocqueetal.,2001),asitwasperformedattheotherscales(Hendrickxetal.,2001).Theserologicalprevalenceofthedisease(prevalenceofantibodiesdirectedagainsttrypanosomalantigens)wasstudiedonasampleofcattlefarmsinthestudyarea,tovalidatethelistofepidemiologicalriskareasidentiﬁed.However,thedataobtainedwerebothpartialandspatiallydisjointed.Themethoddescribedherewassubsequentlydevelopedforestimatingandmodellingdiseasespatialdistribution,withaviewtomakingthedatacompatiblewiththelayersofgeographicdataavailableforthestudyzoneasawhole.2.Materialandmethods2.1.StudyzoneThestudywasconductedinpart(1200km2)oftheSid´eradougouagro-pastoralzonesouthofBobo-Dioulasso(BurkinaFaso),118Nand48W(Fig.1).Thezonehas1000–1100mmofrainfallperyear,withadryseasonfromNovembertoAprilandarainyseasonfromMaytoOctober.ItistypicaloftheSudaniantropicalclimatezone,withbushysavannasandforeststandsalongitswatercourses.Thesetypesofriversidevegetationarethepreferredbiotopesofthetsetseﬂiesfoundinthezone,GlossinatachinoidesandGlossinapalpalisgambiensis(Challier,1973;Gruvel,1975).2.2.Population,samplinganddiagnosisThecattleinthezonewerecountedexhaustively,basedonthedwellingsbywhichtheyarepennedduringthenight(Micheletal.,1999).Foreachdwelling,thenumberofhead,theirwateringpointsattheendofthedryseason,andinformationontranshumancewererecorded.Thegeographicpositionsofeachdwellingandthewateringpointorpoints(twoatmost)weredeterminedbyglobalpositioningsystem(GPS)(GarminTM).Over800dwellings,with16,576head,werevisited.Theherdsweresplitintothreecategories:(i)smallunitswithoneortwopairsofdraughtoxen;(ii)mixedunits,generallywithfewerthan20head,includingdraughtoxenandafewbreeders;(iii)largeherdsofseveraldozenhead,withtranshumanceoftenpractisedduringthedryseason.Inthiszone,wherelivestockareamajorcomponentoffarmingsystemspractised,therearemanysmall

.giF.1coLnoitafoehtRocqueetal.,2001).sydut.enozhTeSide´radougousap-orgalarotenozsidetacolnihtehtuosfoossaluoiD-oboBanikruB(saF,)ota11N8nadW84retfa(DeaL

4757677787970818283848586878889809192939495969791013011100546017018014J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–14Table1HerdsizeandheadnumberinthelocalpopulationHerdsizeUnderfivehead5–20headOver20headNNuummbbeerrooffahneirdmsals1347726((959%%))1816881((2141%%))13133473((1870%%))latoT16587061andmedium-sizedherds,whichaccountforover80%offarmsbutonly20%oftheanimals.Ontheotherhand,80%ofthecattleinthezoneareownedby17%ofthefarmers(Table1).Theherdsarefoundinthreemainzones(Fig.2):(i)anagriculturalzoneinwhichanimalproductioniscloselyintegratedintothefarmingsystem,withmedium-sizedherds,intheeast(zone1);(ii)amixedagriculturalandpastoralzoneinthewest,withsmallandlargeherds(zone2);(iii)analmostexclusivelypastoralzoneinthesouth,withlargeherds(zone3).Thisdistributioncorrespondstothepatternforcrops(DeLaRocqueetal.,2001).Inthewholestudyarea,thereareveryfewtradingandnon-tradingexchangesofcattle.Atwo-stagesamplingwasperformed.Theﬁrstsamplingunitwasonherd,i.e.ananimalmanagementunitsubjecttocommonanimalproductionpractices.Itwaseasilyidentiﬁableintheﬁeldandcorrespondstoanepidemiologicalentity.Theherdsweredrawnatrandom.Thesecondsamplingunitwasanimalswhichwerechosenasfollows:(i)exhaustivesamplinginsmallherds(fewerthanﬁvehead);(ii)10headatmostinmedium-sizedherds(betweenﬁveand20head);(iii)20headatmostinlargeherds(over20head).Withintheherds,theheadweredrawnatrandom,withoutreplacement.Forlogisticalreasonsitwasdecidedtosample2000headspreadover15%oftheherdsinthezone.Aquestionnaireonanimalproductionpracticeswasﬁlledinforeachherd.Bloodsamplesweretakenfromthejugularvein.TheplasmawasanalysedinthelaboratoryusingthreeindirectELISAsystems(T.vivax,T.bruceiandT.congolense),revealingantibodiesagainstTrypanosomaspp.(Desquesnesetal.,2000).2.3.AvailabledataandstatisticalmodelSeveraltypesofdatawereusedtoanalyseandmodeltrypanosomosisseroprevalenceintheherds:Serologicaldatacorrespondingtothevariabletobeexplained.Animalhusbandrydataobtainedfromthefieldsurvey:herdsize,transhumancepracticesandthetypeofwateringpointusedattheendofthedryseason.SpatialdatageneratedbytheGISfromthegeographicpositionofthedifferentunits:distancebetweendwellingandwateringpoint,andproximityofdwellingstothehydrologicalnetwork.Thedescriptivevariableswereclassiﬁedaccordingtoknowledgeofpractices(Lhosteetal.,1993),andtheirepidemiologicalsigniﬁcance(Table2).Thetypeofwateringpointwasdividedintospringsandrivers(whicharepropitioustotsetseﬂies),andwellsand

901011111211311411511611711811911021121221J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–145Fig.2.Herds,sampledherdsandagriculturaldistributioninthestudyzone.Thesizeofthepointsvariesaccordingtothelogeofherdsize.Theherdsarefoundinthreemainzoneswhicharedelineatedbythehydrographicnetwork:(i)anagriculturalzoneinwhichanimalproductioniscloselyintegratedintothefarmingsystem,withmedium-sizedherds,intheeast;(ii)amixedagriculturalandpastoralzoneinthewest,withsmallandlargeherds;(iii)analmostexclusivelypastoralzoneinthesouth,withlargeherds.Thisdistributioncorrespondstothepatternforcrops.boreholes(whicharegenerallyfoundinzonesnotfavourabletotheﬂies).Thezoneclassedasneighbouringonthehydrographicnetworkwassetat2km,basedonknowndataonthetsetseﬂy’sabilitytospread(Cuisanceetal.,1985).Theserologicalprevalenceforeachherdwasmodelledusinglogisticregression,sincetheresponsevariableisaproportionandtheerrorfunctionisassumedtofollowthebinomiallaw(McCullaghandNelder,1989).Thelinkfunctionusedwasthelogitfunction,deﬁnedaslogitðpÞ¼logeðp=ð1pÞÞ.Anover-dispersionphenomenonoftenappearsinusinggen-eralizedlinearmodelwithalogitlinkwhentheresponsevariableisaproportion.Over-dispersionmeansthatthevarianceoftheresponsevariableexceedsthebinomialvarianceandthisproblemisverycommoninlarge-scaleepidemiologicalstudies(McCullaghandNelder,1989).Takingintoaccounttheover-dispersionproblem,weusedaquasi-likelihoodapproach(McCullaghandNelder,1989)intheplaceoflikelihoodfunction.Thisledtowiderconﬁdenceintervalofparametersthantheclassicalapproach.Forthesamereasons,totestthecontributionofthedifferentdescriptivevariablesinthemodel,weconducteda

3214215216217218219210311312313314315316317318319310411412413414415416417411844196J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–14Table2DescriptivevariablesusedformodellingaCodeVariableLevelstypewTypeofwateringpointusedindryseasonArt:artificialNat:naturaldistwDistancebetweenfarmandwateringpointusedindryseason1:<1000m2:1000–4000m3:>4000m1:small(<5)2:medium(5–20)3:large(>20)oNseYoNseYhrdszHerdsizeHy2kmFarmlessthan2kmfromhydrographicnetworkallyrAnimalskeptbydwellingallyearroundaReferencelevelsforthemodelareshowninitalic.devianceanalysiswithF-testintheplaceofw2-test(Collet,1991).Thecoefﬁcientsobtainedinthemodelwereinterpretedbycalculatingtheoddsratiosandtheirconﬁdenceinterval(Bouyeretal.,1995).Thisenabledustoquantifytheriskfactorsassociatedwiththelevelsofeachoftheexplanatoryvariablesinrelationtoareferencelevel.Asthemodelwasnotspatialized,itwasnecessarytolookforautocorrelationamongresiduals.Ifthepresenceofautocorrelationwasdetected,itcouldimplytheomissionofregressorvariables,thepresenceofnon-linearrelationshipsorthattheregressionmodelshouldhaveanautoregressivestructure(CliffandOrd,1973).Totesttheautocorrelation,weﬁrstlyestablishedneighbourhoodrelationshipsbetweenherdsusingaDelaunaytriangulationasproposedbySchmoyer(1994).Inthesecondstep,Geary(Geary,1954)andMoran’s(Moran,1948)statistics(seealsoCliffandOrd,1973)werecomputedforresiduals.Bypermutingthevaluesoftheresidualmap,wecomputednewvaluesofautocorrelationstatisticsandtheobservedvalueistestedbycomparingtothesetofvaluesobtainedforthepermutations.Asthenumberofpossiblepermutationwasverylarge,weusedaMonte-Carlo(Manly,1991)versionofthetest.Thesameprocedurehasbeencarriedoutforobservedandpredictedprevalencevalues.ThiskindofprocedurehasbeenrecentlyusedinKleinschmidtetal.(2000)withthenon-parametricD-statistic(Walter,1992)tomeasureautocorrelationofpredictionsfromalogisticregression.Whentheseindiceswereappliedtoobservedandpredictedprevalencevaluesandtheresidualsofthemodel,theyenabledustotestthecapacityofamodeltotakeaccountofthespatialnatureofdata.Thestatisticalmodelwastheninvertedtoestimatetheserologicalprevalenceforalltheherdsinthezone,usingtheexplanatoryvariablessharedwiththesurvey,whichwerethesamethanthoseusedtogeneratethemodel.AllcalculationsweremadeusingtheRsoftware(IhakaandGentleman,1996).2.4.SpatialmodelSpeciﬁcproblemslinkedtothegeographicalnatureofmappedobjectssuchasherdsandthe‘‘inherd’’variabilityofmeasuredvariableshavetobeconsideredwhenmapping

051151251351451551651751851951061161261361461561661761861961071171271371471571671771871971081181281381481581681781881819091J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–147seroprevalence:(i)herdsarepointsthatmaybesuperimposediftheyareclosetooneanother,hencemaskinginformation;(ii)themeaningofprevalencewithinaherdvarieswiththenumberofheadintheherd,aprevalenceof50%inatwoheadherdhasnotthesamesigniﬁcanceasaprevalenceof50%ina100headherd;(iii)mappingonlythepointscorrespondingtothepensusedatnightprovidesonlyapartialrepresentationofrealityasanimalsmoveandoccupyacontinuousspace;(iv)spatialinformationonprevalencehastobecompatiblewiththeotherinformationavailableintheGISiftheyaretobecompared.Toovercometheseproblems,aspatialmodeloflandoccupationbycattleandofdiseasedistributionwasdeveloped.AllspatialobjectmanipulationsusedtheMapinfoTMsoft-.erawTherepresentationoflandoccupationbycattleinazoneasawholeisbasedonmodellingthedailymovementsoftheanimalsineachherd.Insavannazones,wateravailabilityisthemainconstraintattheendofthedryseason,andgovernsmovements(Boutrais,1994).Herdmovementswerethereforemodelledbyrepresentingthedirectroutebetweenthenightpenandthewateringpointorpoints,anddrawingabufferzonearoundtheroute,correspondingtotheareaoccupiedbythecattleduringtheday(Micheletal.,1999).Thiszoneofdailyusebytheherdvariesinsize.Thewidertheherdandtheneareritistoitswateringpoint,thelargerthezoneoffrequentation(Fig.3).Thismodelwasvalidatedbymonitoringthemovementsofasampleofherds.Thepredictedprevalenceforalltheherdswasappliedtotheirzonesofdailyuse.Tosynthesizethisinformation,whichwasnotyetveryeasytoresolveduetothesuper-impositionofpolygons,itprovednecessarytoaggregateitsoastoshifttoasmallerscale.Thiswasdonebyprojectingallthezonesofuseandthecorrespondingprevalencesontoaregulargeographicgridof1km2(Raynaletal.,1996).Thecumulateddistributionofantibody-prevalenceinthestudyzonewasthenrepresentedbyassigningtoeachsquarethemeanvalueoftheprevalencesfortheherdpolygonsimpingingonit,soastoproduceamapofaverageprevalence(Fig.4).Thecalculatedmeanvalueofprevalencewasweightedbythesizeofherdsinordertotakeintoaccountforproblems(ii)citedabove.Smoothingbytwo-dimensionalweightedlocalregression(ClevelandandDevlin,1988)onthecentroidsofthesquaresinthegridmadethemapsmorerealistic.3.Results3.1.SamplingandobservedseroprevalenceIntotal,216herdsand1784headweresampled.Herdandcattledistributioninthesampleshowedthatsmallherdswereslightlyunder-represented,infavourofmedium-sizedherds(Table3).Ontheotherhand,smallherdswereover-representednearthehydrographicnetwork(Fig.2).Thedifferencesinrelationtothesampleinitiallyplannedcanbeattributedtoﬁeldconstraintssuchasherdsinanout-of-the-wayplaceorcattlebreederabsentornotinagreementwithtakingabloodsample.Theaverageserologicalprevalenceobservedamongthecattlewas73.4%.Themapofherdprevalence,shownaspointsaccordingtothecorrespondingdwelling,showedcasedistributionbutwasdifﬁculttointerpret(Fig.5).

8J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–14Fig.3.Modellingofdailyherdmovements.Modellingherdmovementsconsistsinrepresentingthedirectroutebetweenthenightpenandthewateringpointorpoints,anddrawingabufferzonearoundtheroute,correspondingtotheareaoccupiedbythecattleduringtheday.Thiszoneofdailyusebytheherdvariesinsize.Thelargertheherdandtheneareritistoitswateringpoint,thelargerthezoneoffrequentationis(BV:cattle)(afterDeLaRocqueetal.,2001).Table3HerdsizeandheadnumberinthesampleHerdsizeUnderfiveheadNumberofherds110(51%)Numberofanimals327(18%)5–20head70(32%)736(41%)Over20head36(17%)721(41%)latoT1271864

191291391491591691791891991002J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–149Fig.4.Dataaggregationandspatialdistributionofmeanprevalence.Thiswasdonebyprojectingallthezonesofuseandthecorrespondingprevalencesontoaregulargeographicgridof1km2.Thecumulateddistributionofantibody-prevalenceinthestudyzonewasthenrepresentedbyassigningtoeachsquarethemeanvalueoftheprevalencesfortheherdpolygonsimpingingonit,soastoproduceamapofaverageprevalence.3.2.Statisticalmodelling:identificationofriskfactorsThedevianceanalysisshowedthatonlythedistancebetweencattlepenandwateringpointwasnotsigniﬁcantandthisvariablewasexcludedfromthemodel.Alltheothervariablesweresigniﬁcant(Table4).Thedispersionparameterforthemodelwas2.48.Therelationbetweenthenumbersofobservedandpredictedpositives(Spearman’srankcorrelationr¼0:45,P<0:0001)showedthatthestatisticalmodelhasagoodﬁt.Thespatialautocorrelationtestsrevealedapositivecorrelationbetweenobservedandpredictedprevalences,whereastheresidualsofthemodelwerenotcorrelated(Table5).Thevariablesusedinthemodelthustakeaccountofthespatialfactor.TheoddsratioscalculatedwiththecoefﬁcientsestimatedbythemodelshowedthatproximitytotheTable4DevianceanalysisofthemodelDegreeofDevianceResidualdegreeResidualP(>F)freedomoffreedomdevianceNULL––215666.51–typew117.66214648.86<0.0082hrdsz237.37212611.49<0.0006allyr159.07211552.41<0.0001hy2km154.17210498.24<0.0001

10222030240250260270201J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–14Fig.5.Distributionofserologicalprevalenceamongtheherdssampled.Themapofseroprevalencefortheherdssampled,shownaspointsaccordingtothecorrespondingdwelling,indicatedcasedistribution.BV:cattle.hydrographicnetwork,frequentationofnaturalwateringpoints,largeherdsizeandthefactofkeepinganimalsneardwellingsallyearroundwereallriskfactors(Table6).3.3.ModellingofspatialdistributionofprevalenceThemapofpredictedserologicalprevalences,obtainedbyspatialmodellingonawholestudyzonescale,showedthatmeanserologicalprevalenceisdistributedalongthehydrographicnetwork,withfocalpointsofhighvalues,andthatitspreadsradiallyintotheneighbouringsavannas(Fig.6).Table5Spatialautocorrelationtestsfortheobservedandthepredictedprevalencesandtheresiduesofthemodel,usingMoran’s(I)andGeary’s(c)indexesVariableIObservedprevalence0.196Predictedprevalence0.542Residualsofmodel0.020P-values<0.001<0.001842.0c997.0564.00279.P-values<0.001<0.001882.0

802902012112212312412J.-F.Micheletal./PreventiveVeterinaryMedicine1728(2002)1–1411Table6Oddsratios(OR)calculatedfromthecoefficientsoftheserologicalmodelVariableLowerconfidenceORinterval(OR)Intercept0.531.08typewnat0.881.34hrdsz20.380.67hrdsz31.031.99allyryes1.692.70hy2kmyes1.943.43Upperconfidenceinterval(OR)91.240.261.148.364..3017Fig.6.Distributionofmeanserologicalprevalenceinthezoneandhigh-transmission-riskzones.Thehigh-transmissionriskzonesaredelineatedbyawhiteoutline.Thismapshowsthatmeanserologicalprevalenceisstructuredlinearlyalongthehydrographicnetwork,withfocalpointsofhighvalues,andthatitspreadsradiallyintotheneighbouringsavannas.4.Discussion4.1.ObservedprevalenceandstatisticalmodelTheserologicalresultsobtainedfromthesampleconﬁrmtheenzooticsituationthathadalreadybeenobservedfortrypanosomosisintheSide´radougouzone,withhighinfectionlevelsamongvectors(DeLaRocque,1997).Onananimalproductionzonescale,parasitepressurecanbeevaluatedmoreaccuratelybythenumberofantibodycarriersthanbydirectdetectionofparasites(Desquesnesetal.,2000).Withserologicaldata,thestatisticalmodel

Voir