[R] Anova und Tukey HSD
Meyners, Michael
meyners.m at pg.com
Mon May 13 17:24:04 CEST 2013
Jochen,
a) this is an English-spoken mailing list; other languages are not encouraged nor will they typically generate a lot of replies...
b) your code is fine, so this is not an R-issue; you are rather stuck with some of the stats background -- you might want to see a friendly local statistician for advice :-) or try a stats mailing list (see e.g. the posting guide)
c) your interpretation is ok -- if the CI does not contain 0, you'd have significance at the respective level. Alternatively, you might also use the p value (column "p adj") and compare that to your significance level (and it seems odd that any search machine might fail to give you some links on the relationship of p values and CIs -- I got plenty at first try)
d) In brief: you are doing ~100 (rough estimate) paired comparisons. With that, you can expect TukeyHSD to be pretty conservative, and it does not come as a surprise that you are unable to identify pairs of different groups, even less as the p value from the F test is not extremely small (~3%; whatever "extremely small" may mean in this case). Google also easily gets me to this site (and many other useful ones): http://www.pmean.com/05/TukeyTest.html which might provide some further guidance.
Net, tough luck to encounter such a situation, but apparently none of the group differences is strong enough to still show up as "significant" after correcting for multiplicity (which is certainly called for) in this case where you have an inflated number of significance tests.
HTH, Michael
> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On Behalf Of Jochen Schreiber
> Sent: Montag, 13. Mai 2013 14:25
> To: r-help at r-project.org
> Subject: [R] Anova und Tukey HSD
>
> Hallo zusammen,
>
> ich mache zuerst mittels eine Anova eine Überprüfung ob in den Daten ein
> signifikanter Unterschied exisitiert. Danach will ich mittels des TukeyHSD
> rausfinden, zwischen welchen Gruppe der Unterschied vorliegt. Allerdings
> weiß ich nicht wie ich die Daten interpretieren soll. Hier erstmal mein R Code:
>
> [code]
> Value<-c(-0.9944999814033508,-0.35850000381469727,0.7063000202178955,-
> 1.774399995803833,-
> 1.080299973487854,0.30550000071525574,1.8499999046325684,-
> 0.4124999940395355,0.5827999711036682,1.7506999969482422,-
> 6.693999767303467,-0.8779000043869019,-
> 1.3408000469207764,1.2560999393463135,-
> 0.10040000081062317,1.8499999046325684,-
> 0.3319000005722046,0.4957999885082245,0.8779000043869019,0.73879998922
> 34802,0.8779000043869019,0.9154000282287598,0.8779000043869019,0.70630
> 00202178955,-1.3408000469207764,0.7063000202178955,-
> 0.3319000005722046,-1.6448999643325806,0.4124999940395355,-
> 1.6448999643325806,-
> 0.8779000043869019,0.7487000226974487,0.4399000108242035,1.84999990463
> 25684,-1.6448999643325806,-2.4323999881744385,1.2265000343322754,-
> 0.4957999885082245,-9.999899864196777,-1.7506999969482422,-
> 1.6448999643325806,-9.999899864196777,0.8779000043869019,-
> 5.06279993057251,0.8779000043869019,-2.9677000045776367,-
> 5.06279993057251,-6.693999767303467,-
> 1.0990500450134277,0.9944999814033508,-0.4677000045776367,-
> 0.35850000381469727,-
> 9.999899864196777,0.5827999711036682,0.7487000226974487,0.738799989223
> 4802,-0.2533000111579895,-9.999899864196777,-
> 1.0363999605178833,0.30550000071525574,-1.1749999523162842,-
> 0.8064000010490417,-9.999899864196777,-0.9944999814033508,-
> 2.478300094604492,-0.1509999930858612,0.4957999885082245,-
> 4.571800231933594,-6.324900150299072,-0.38530001044273376,-
> 1.3408000469207764,-5.93179988861084,-6.693999767303467,-
> 2.9677000045776367,0.8779000043869019,-0.050200000405311584,-
> 1.774399995803833,-0.1509999930858612,-0.23725003004074097,-
> 0.6432999968528748,1.2560999393463135,-
> 0.10040000081062317,0.4399000108242035,-
> 0.7063000202178955,0.9154000282287598,-
> 0.21819999814033508,1.2265000343322754,-
> 0.4124999940395355,0.17640000581741333,-1.4758000373840332,-
> 0.9944999814033508,-1.080299973487854,-0.6432999968528748,-
> 9.999899864196777,-2.0536999702453613,-
> 0.21819999814033508,0.7487000226974487,0.025100000202655792,-
> 1.0363999605178833,-0.050200000405311584,-
> 0.7387999892234802,0.4957999885082245,-1.4758000373840332,-
> 0.7063000202178955,0.17640000581741333,-5.06279993057251,-
> 0.6432999968528748,-1.4758000373840332,-0.9944999814033508,-
> 0.2533000111579895,0.17640000581741333,-
> 0.3319000005722046,0.6776500344276428,0.30550000071525574,-
> 0.050200000405311584,0.5827999711036682,1.2560999393463135,-
> 0.4957999885082245,-0.38530001044273376,0.9944999814033508,-
> 2.4323999881744385,1.1263999938964844,-
> 0.9944999814033508,1.7506999969482422,1.080299973487854,-
> 0.7387999892234802,-1.3408000469207764,0.6128000020980835,-
> 2.0536999702453613,0.7063000202178955,-0.8064000010490417,-
> 0.8779000043869019,-0.050200000405311584,-2.9677000045776367,-
> 0.8779000043869019,-2.0536999702453613,-1.3408000469207764,-
> 1.3408000469207764,-0.38530001044273376,0.7063000202178955,-
> 9.999899864196777,-
> 0.4677000045776367,0.7721999883651733,0.025100000202655792,1.126399993
> 8964844,-6.324900150299072,-0.1509999930858612,-0.4399000108242035,-
> 0.9944999814033508,-0.9944999814033508,-0.4677000045776367,-
> 1.0363999605178833,-1.7506999969482422,1.2265000343322754,-
> 0.8779000043869019,0.6128000020980835,-
> 0.050200000405311584,0.5827999711036682,-0.7063000202178955,-
> 0.6432999968528748,-
> 0.23725003004074097,0.025100000202655792,0.4124999940395355,0.77219998
> 83651733,-1.0990500450134277)
> Value<-c(Value,0.9944999814033508,-
> 0.2533000111579895,1.2560999393463135,-0.21819999814033508,-
> 1.1749999523162842,-0.38530001044273376,-0.4399000108242035,-
> 0.7063000202178955,-2.478300094604492,-
> 2.4323999881744385,0.9154000282287598,-0.23725003004074097,-
> 0.38530001044273376,-1.6448999643325806,-
> 0.050200000405311584,1.8499999046325684,-0.38530001044273376,-
> 0.6432999968528748,-4.571800231933594,-6.693999767303467,-
> 1.7506999969482422,1.080299973487854,0.4124999940395355,-
> 1.3408000469207764,-5.93179988861084,-0.35850000381469727,-
> 0.6432999968528748,-0.4124999940395355,-1.0990500450134277,-
> 0.9944999814033508,-0.8064000010490417)
> Group<-
> factor(c(rep('D',18),rep('C',1),rep('A',7),rep('B',34),rep('E',3),rep('F',4),rep('G'
> ,10),rep('H',2),rep('I',29),rep('J',16),rep('N',1),rep('M',1),rep('Z',2),rep('X',67),
> rep('O',1)))
> mydata<-data.frame(Group, Value)
> summary(aov(Value ~Group,mydata))
> TukeyHSD(aov(Value ~Group))
> [/code]
>
> Das Summary von Anova gibt folgenden Output:
>
> [code]
> Df Sum Sq Mean Sq F value Pr(>F)
> Group 14 147.47 10.5339 1.8815 0.03089 *
> Residuals 181 1013.38 5.5988
> ---
> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 [/code]
>
> Dort sehe ich das der Pr(>F)-Wert unter 0.05 liegt daher existiert ein
> signifikanter Unterschied. Nun folgt ein Teil der Ausgabe vom TukeyHSD:
>
> [code]
> $Group
> diff lwr upr p adj
> B-A -2.42876615 -5.8061007 0.9485684 0.4650621 C-A 0.38050000 -8.3184455
> 9.0794455 1.0000000 D-A -0.78451112 -4.4090717 2.8400495 0.9999881 E-A -
> 4.49116661 -10.1063118 1.1239785 0.2838741 F-A -1.27940002 -6.3796089
> 3.8208088 0.9999272 G-A -3.41370001 -7.4237157 0.5963157 0.1940527 H-A -
> 0.69152501 -7.2157341 5.8326841 1.0000000 I-A -1.33448620 -4.7611687
> 2.0921963 0.9910595 J-A -0.43914063 -4.1265855 3.2483043 1.0000000 M-A -
> 1.23619999 -9.9351455 7.4627455 0.9999999 N-A 0.58289997 -8.1160455
> 9.2818454 1.0000000 O-A -1.30380000 -10.0027455 7.3951455 0.9999999 X-A -
> 1.40412389 -4.6363376 1.8280899 0.9756749 Z-A -0.86140003 -7.3856091
> 5.6628091 1.0000000 C-B 2.80926615 -5.4466486 11.0651809 0.9977027 D-B
> 1.64425503 -0.7276462 4.0161562 0.5299872 E-B -2.06240046 -6.9632505
> 2.8384496 0.9816337 F-B 1.14936613 -3.1518670 5.4505992 0.9998458 G-B -
> 0.98493386 -3.9121684 1.9423006 0.9979595 H-B 1.73724114 -4.1833822
> 7.6578644 0.9995457 I-B 1.09427995 -0.9625693 3.1511291 0.8829284 J-B
> 1.98962552 -0.4773008 4.4565519 0.2709375 M-B 1.19256616 -7.0633486
> 9.4484809 0.9999999 N-B 3.01166612 -5.2442486 11.2675808 0.9952989 O-B
> 1.12496615 -7.1309486 9.3808809 1.0000000 X-B 1.02464226 -0.6887408
> 2.7380253 0.7612648 [/code]
>
> Wie sehe ich nun welche Gruppe/n den signifikanten Unterschied
> aufweisen? Ich meine mich zu erinnern (habe aber leider nix dergleichen
> gefunden), dass wenn das Intervall die 0 nicht mit einschließt ein
> signifikanter Unterschied besteht. Was bedeuteten würde, wenn der pwr
> und lwr > 0.0 oder pwr und lwr < 0.0 wären das dann ein signifikanter
> Unterschied da wäre. Nur habe ich das in keiner Zeile der TukeyHSD
> Ausgabe?
>
> Ich denke das ist ne Kleinigkeit aber ich hoffe ihr könnt mir helfen.
>
> Viele Grüße
>
> bladepit
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-
> guide.html
> and provide commented, minimal, self-contained, reproducible code.
More information about the R-help
mailing list