Fishers Least Significant Difference (LSD) test in Prism

Fishers Least Significant Difference (LSD) test in Prism. Following one-way (or two-way) analysis of variance (ANOVA), you may want to explore further and ... PleaseenableJavaScripttoviewthissite. StatisticsGuide CurveFittingGuide PrismGuide Resources FreeTrial ZoomWindowOut LargerText | SmallerText HidePageHeader ShowExpandingText PrintableVersion SavePermalinkURL Navigation:PRINCIPLESOFSTATISTICS>Multiplecomparisons>Threeapproachestodealingwithmultiplecomparisons>Approach1:Don'tcorrectformultiplecomparisons Fisher'sLeastSignificantDifference(LSD) Scroll Prev Top Next More FishersLeastSignificantDifference(LSD)testinPrism Followingone-way(ortwo-way)analysisofvariance(ANOVA),youmaywanttoexplorefurtherandcomparethemeanofonegroupwiththemeanofanother.Onewaytodothisisbyusing Fisher'sLeastSignificantDifference(LSD)test. KeyfactsaboutFisher'sLSDtest •TheFishersLSDtestisbasicallyasetofindividualttests.ItisonlyusedasafollowuptoANOVA.•UnliketheBonferroni,Tukey,DunnettandHolmmethods,Fisher'sLSDdoesnotcorrectformultiplecomparisons.•IfyouchoosetousetheFisher'sLSDtest,you'llneedtoaccountformultiplecomparisonswhenyouinterpretthedata,sincethecomputationsthemselvesdonotcorrectformultiplecomparisons.•TheonlydifferencebetweenasetofttestsandtheFisher'sLSDtest,isthatttestscomputethepooledSDfromonlythetwogroupsbeingcompared,whiletheFisher'sLSDtestcomputesthepooledSDfromallthegroups(whichgainspowerbutdependsontheassumptionthatallgroupsaresampledfrompopulationswiththesameSD).•PrismperformstheunprotectedLSDtest.Unprotectedsimplymeansthatcalculationsarereportedregardlessoftheresultsofthe ANOVA.TheunprotectedFisher'sLSDtestisessentiallyasetofttests,withoutanycorrectionformultiplecomparisons.•PrismdoesnotperformaprotectedFisher'sLSDtest.ProtectionmeansthatyouonlyperformthecalculationsdescribedabovewhentheoverallANOVAresultedinaPvaluelessthan0.05(orsomeothervaluesetinadvance).Thisfirststepsortofcontrolsthefalsepositiveratefortheentirefamilyofcomparisons.WhiletheprotectedFisher'sLSDtestisofhistoricalinterestasthefirstmultiplecomparisonstesteverdeveloped,itisnolongerrecommended.Itpretendstocorrectformultiplecomparisons,butdoesn'tdosoverywell.•Howitworks.  ©1995-2019GraphPadSoftware,LLC.Allrightsreserved.