ThePigITProjectaimsatimprovingwelfareandproductivityofslaughterpigsbyintegrationofvarious
sensorsystemsforalarmpurposes.Herewepresentanexploratoryanalysistoassessthepredictivevalueof
temperaturesensordatawithrespecttopenfoulinganddiarrhea.Werecordedthetemperatureattwo
locationsin8pensbetweenNovember2013andDecember2014.Asinglelogisticregressionmodelwas
madetoexpresstheprobabilityofeitherdiarrheaorfoulingperpenperday,andwasreducedviaback-
wardselimination.Thepredictiveperformanceswereevaluatedbytheareaunderthereceiveroperating
characteristicscurve(AUC).IndiscriminantpredictionofeithereventreachedanAUCof0.80.Similarper-
formanceswereseenwhenpredictingeachoftheeventsontheirownusingthesamemodel,withAUC
valuesat0.78and0.81fordiarrheaandfouling,respectively.Thus,temperatureinformationseemsto
providepredictivevalueinrelationtofoulinganddiarrhea.Itwouldbemeaningfultocombinethisin-
formationwithotheravailabledatabyusingmoreadvancedmodelstoachieveanoptimalpredictivepower.