多重比較Multiple comparisons @ 凝視、散記 - 隨意窩
文章推薦指數: 80 %
如果要作post hoc comparisons (有人叫posteriori test) ,有許多選擇:1) Fisher's Least Significant Difference (又稱為LSD);2) Tukey's Test;3) The Ryan Procedure ...
凝視、散記【散點透視】APersonalCasualNotesconsistofScatteredPerspectivesforLearning,ThinkingandPractices日誌相簿影音好友名片
202011081030多重比較Multiplecomparisons?Insight&Algorithm在用ANOVA時,會得到一個F-test,這個test 只能告訴你整個model是否顯著(i.e.,總體平均數不相同或不完全相同),並不能告訴你特定的組與組之間是否有差異。
有很多種方法可以比較組間的平均值是否有差異,這些方式就稱作多重比較(multiplecomparisons)
如果要作posthoccomparisons(有人叫posterioritest),有許多選擇:1)Fisher’sLeastSignificantDifference(又稱為LSD);2)Tukey’sTest;3)TheRyanProcedure(REGWQ);4)TheSchefféTest;5)Dunnett’stestforcomparingalltreatmentswithacontrol。
會產生這麼多方法,主要是比較的方式不同。
有的方法是與均值作比較 (i.e.,所有組別的平均皆相同),有的方法是配對比較 (pairwisecomparison)(i.e.,與控制組比較、與最佳組比較),有的是contrast(特定兩組比較)。
比較表 (出自StatisticalmethodsforPsychologybyDavidC.Howell,第六版,頁375):
Test
Errorrate
Comparison
Type
事前/事後
備註
Individual t tests
PCa
Pairwise
t
事前
Linearcontrast
PC
Anycontrasts
F
事前
Bonferroni t
FWb
Anycontrasts
t**
事前
Holm:Larzelere&Mulaik
FWb
Anycontrasts
t**
兩者
Fisher’sLSD
FW*
Pairwise
t
事後
Newman-Keulstest
FW*
Pairwise
Range
事後
爭議性大
Ryan(REGWQ)
FW
Pairwise
Range
事後
TukeyHSD
FW
Pairwise***
Range
事後
只想測試某兩組時
SchefféTest
FW
Anycontrasts
F**
事後
想測試所有差異時
Dunnett’stest
FW
Withcontrol
F**
事後
Note:a:Errorratepercomparison.b:Familyerrorrate(FW). FW*:againstcompletenullhypothesis.t** :modified t test.F*:modified.Pairwise***:TukeyHSDcanbeusedforallcontrasts,butispoorforthispurpose.
----------------------
AfteryouhaverunANOVAandgotasignificantF-statistic(i.e.youhaverejectedthenullhypothesisthatthemeansarethesame),thenyourunSheffe’stesttofindoutwhichpairsofmeansaresignificant.TheScheffetestcorrectsalphaforsimpleandcomplexmeancomparisons.Complexmeancomparisonsinvolvecomparingmorethanonepairofmeanssimultaneously.
Outofthethreemeancomparisonstestyoucanrun(theothertwoareFisher’sLSDandTukey’sHSD).TheScheffetestisthemostflexible,butitisalsothetestwiththeloweststatisticalpower.Decidingwhichtesttorunlargelydependsonwhatcomparisonsyou’reinterestedin:
Ifyouonlywanttomakepairwisecomparisons,runtheTukeyprocedurebecauseitwillhaveanarrowerconfidenceinterval.Ifyouwanttocompareallpossiblesimpleandcomplexpairsofmeans,runtheScheffetestasitwillhaveanarrowerconfidenceinterval.
UnderstandingScheffé'sTest
WhileScheffé'stesthastheadvantageofgivingtheexperimentertheflexibilitytotestanycomparisonsthatappearinteresting,thedrawbackofthisflexibilityisthatthetesthasrelativelylowerstatisticalpowerthanteststhataredesignedforpre-plannedcomparisons.
TheScheffétestisoneoftheoldestmultiplecomparisonproceduresinusetoday.ItisimportanttorecognizethatitisafrequentlymisusedprocedureandthatitisalsoavaluabletestwhenusedasHenrySchefféintendedit.UnlikecompetitorssuchasTukey'sHonestlySignificantDifferencetest,theScheffétestisspecificallydesignedforthesituationinwhichposthoccomparisonsinvolvemorethanpairwisedifferences.Forexample,itcouldbeusedtocomparethemeanoftwogroupstothemeanoftwoothergroupsonthebasisofinterestingdifferencesthatappearedafterthedatahadbeencollected.
SchefféTest(雪費檢定;同步檢定法) post-hoctest
同時測試所有可能的contrast,並視是否其中之一為0。
臨界值最大,最不容易顯著(也表示較不容易犯TypeIIerror)。
可用於各組人數不同、非常態分配上。
Whatarethemajordifferencesbetweenscheffeandtukeyposthoctests?
Whatarethekeyconditionsthatnecessitatetheuseofeitherscheffeandtukeyposthoctest?
Generally,TukeyandScheffétestsaremoreconservative.Theyfindithardertoseedifferencesandgenerallygivethesameresult.Inrelationtothedifferences:-Inpairwisecomparisons,TukeytestisbasedonstudentizedrangedistributionwhileScheffeisbasedinFdistribution.-Tukey'stestisveryrigorous,controllingthetypeIerrorverywell,butfavorsthetypeIIerror.-TheScheffetestallowscomparinganycontrastbetweenmeansandallowsdifferentnumberofobservationspertreatment.
OR
Wewouldpositthattherearenomajordifferences.Huberty,C.J,&Morris,J.D.(1988).Asinglecontrasttestprocedure.EducationalandPsychologicalMeasurement,48,567-578.
【出處】
https://researcher20.com/2010/05/27/%E5%A4%9A%E9%87%8D%E6%AF%94%E8%BC%83-multiple-comparisons/
https://www.statisticshowto.com/scheffe-test/
https://www.investopedia.com/terms/s/scheffes-test.asp
https://www.researchgate.net/post/What_are_the_major_differences_between_scheffe_and_tukey_post_hoc_testsJason/Xuite日誌/回應(0)/引用(0)Whatisthedif...|日誌首頁|Journey-Don't...上一篇Whatisthedifferencebetweenlexicalandsemantic?...下一篇Journey-Don'tStopBelievin'(整晚只為一笑,不顧一切賭一把)...回應
全部展開|全部收合
全部展開|全部收合
累積|今日loading......
關鍵字
即時新台幣匯率
匯率
COMMODITIES
延伸文章資訊
- 1多重比較
❏ 屬於試探性資料分析Exploratory data analysis -- Tukey, 1977. ❏. 11. Page 12. 事後比較. ❏ 實際的 ...
- 2多重比較Multiple comparisons @ 凝視、散記 - 隨意窩
如果要作post hoc comparisons (有人叫posteriori test) ,有許多選擇:1) Fisher's Least Significant Difference (又稱...
- 3多重比較Multiple comparisons - 研究生2.0
Tukey HSD 是測試所有可能的pairwise 的平均值,並決定是否其中之一為0。 Tukey's test 的假設是1) 所有的observation 都是獨立的;2) 平均數是來自常態...
- 4如何以正確的方式來進行多重對比測試? - BSE Lab 首頁
執行MCT的方法有幾種,例如Tukey方法,Newman- Keuls 方法,Bonferroni 方法,Dunnett 方法,Scheffé 檢定等。在本文中,我們討論了用於分析給定數據的最佳...
- 5什么是多重比较的Tukey 法? - Support - Minitab
Tukey 法在方差分析中用于在将全族误差率控制在指定水平的同时,为因子水平均值之间的所有配对差异创建置信区间。进行多重比较时考虑全族误差率很重要,因为对于一系列 ...