C. Weiner
2012-Oct-07 23:37 UTC
[R] Why do I get different results for type III anova using the drop1 or Anova command?
Dear experts,
I just noticed that I get different results conducting type III anova
using drop1 or the Anova command from the car package. I suppose I made
a mistake and hope you can offer me some help. I have no idea where I
got wrong and would be very grateful for explaination as R is new
terrain for me.
If I run the commands in line, they produce the same results. But if I
run them in seperate windows, they produce different results.
What I did in easy words:
Opened R prompt, ran parts 0 and 1-> got result 1
Opened new R prompt, ran parts 0 and 2 ->got result 2, ran part 1
following part 2, got result 2a
Why are the results different when they are run seperately and why do
the commands produce the same results when running one after the other?
Which results can I rely on? I based my script on:
http://mcfromnz.wordpress.com/2011/03/0 ... explained/
<http://mcfromnz.wordpress.com/2011/03/02/anova-type-iiiiii-ss-explained/>
but I surely made mistakes.
Thanks for help.
Kind regards,
Christiane
#0
library(car)
P<-read.table("rhoPgegenrhoAgewichtet-mit
Befruchtungssystem0912.txt",
sep="\t",dec=",", header=T,encoding="latin1")
summary(P)
str(P)
#1
modD3 <- lm(P$rhoPug ~ P$rhoAg*P$dprime*P$Befruchtungssystem, data=P)
Anova(modD3,contrasts=list(rhoAg=contr.sum, dprime=contr.poly),
type="III") ###liefert dieses Ergebnis, wenn der Befehl als erster
l?uft
#Response: P$rhoPug
# Sum Sq Df F value Pr(>F)
#(Intercept) 0.07719 1 4.5075 0.03530 *
#P$rhoAg 0.00687 1 0.4014 0.52730
#P$dprime 0.00176 1 0.1029 0.74882
#P$Befruchtungssystem 0.00250 1 0.1461 0.70280
#P$rhoAg:P$dprime 0.00931 1 0.5439 0.46193
#P$rhoAg:P$Befruchtungssystem 0.00001 1 0.0004 0.98343
#P$dprime:P$Befruchtungssystem 0.00108 1 0.0630 0.80214
#P$rhoAg:P$dprime:P$Befruchtungssystem 0.00206 1 0.1205 0.72893
#Residuals 2.70584 158
#2
options(contrasts = c("contr.sum","contr.poly")) #liefert
ein anderes
Ergebnis als der Befehl oben, wenn der obere Befehl NICHT gelaufen ist
modD3 <- lm(P$rhoPug ~ P$rhoAg*P$dprime*P$Befruchtungssystem, data=P)
drop1(modD3,.~.,test="F") # type III SS and F Test
P$rhoPug ~ P$rhoAg * P$dprime * P$Befruchtungssystem
Df Sum of Sq RSS AIC F value Pr(F)
<none> 2.7058 -667.35
P$rhoAg 1 0.0131973 2.7190 -668.54 0.7706 0.3814 !!
P$dprime 1 0.0002240 2.7061 -669.34 0.0131 0.9091 !!
P$Befruchtungssystem 1 0.0025022 2.7083 -669.20 0.1461 0.7028 !!
P$rhoAg:P$dprime 1 0.0283677 2.7342 -667.62 1.6565 0.2000 !!
P$rhoAg:P$Befruchtungssystem 1 0.0000074 2.7058 -669.35 0.0004 0.9834
P$dprime:P$Befruchtungssystem 1 0.0010790 2.7069 -669.29 0.0630 0.8021
P$rhoAg:P$dprime:P$Befruchtungssystem 1 0.0020641 2.7079 -669.22 0.1205
0.7289
#2a (command is the same as in 1, but results are different if the
command is run after the one above)
Anova(modD3,contrasts=list(Befruchtungssystem="contr.sum",
dprime="contr.poly"), type="III") #liefert in nachstehend
nach dem
2.Befehl dieses Ergebnis
#Response: P$rhoPug #also jetzt dasselbe wie drop1, aber nicht dasselbe,
das
#der exakt gleich Befehl alleine geliefert h?tte. Was
#stimmt den nun? Wo liegt mein Fehler?
# Sum Sq Df F value Pr(>F)
#(Intercept) 0.06877 1 4.0158 0.04679 *
#P$rhoAg 0.01320 1 0.7706 0.38136
#P$dprime 0.00022 1 0.0131 0.90909
#P$Befruchtungssystem 0.00250 1 0.1461 0.70280
#P$rhoAg:P$dprime 0.02837 1 1.6565 0.19997
#P$rhoAg:P$Befruchtungssystem 0.00001 1 0.0004 0.98343
#P$dprime:P$Befruchtungssystem 0.00108 1 0.0630 0.80214
#P$rhoAg:P$dprime:P$Befruchtungssystem 0.00206 1 0.1205 0.72893
#Residuals 2.70584 158
John Fox
2012-Oct-08 13:08 UTC
[R] Why do I get different results for type III anova using the drop1 or Anova command?
Dear Christiane, I found the code and output that you attached difficult to read, but I see at least one error: I'm not sure why you think that Anova() takes a contrasts argument. It doesn't; contrasts is an argument to lm(). See ?Anova and ?lm. Here's a reproducible example that shows that you get the same tests from Anova() and drop1(): ---------------- snip ------------> # proper "type III" tests: > mod.1 <- lm(conformity ~ fcategory*partner.status, data=Moore,+ contrasts=list(fcategory=contr.poly, partner.status=contr.sum))> drop1(mod.1, .~.,test="F")Single term deletions Model: conformity ~ fcategory * partner.status Df Sum of Sq RSS AIC F value Pr(>F) <none> 817.76 142.50 fcategory 2 36.019 853.78 140.44 0.8589 0.431492 partner.status 1 239.562 1057.33 152.06 11.4250 0.001657 ** fcategory:partner.status 2 175.489 993.25 147.25 4.1846 0.022572 * --- Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1> Anova(mod.1, type=3)Anova Table (Type III tests) Response: conformity Sum Sq Df F value Pr(>F) (Intercept) 5752.8 1 274.3592 < 2.2e-16 *** fcategory 36.0 2 0.8589 0.431492 partner.status 239.6 1 11.4250 0.001657 ** fcategory:partner.status 175.5 2 4.1846 0.022572 * Residuals 817.8 39 --- Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1> # hypotheses not sensible: > mod.2 <- lm(conformity ~ fcategory*partner.status, data=Moore) > drop1(mod.2, .~.,test="F")Single term deletions Model: conformity ~ fcategory * partner.status Df Sum of Sq RSS AIC F value Pr(>F) <none> 817.76 142.50 fcategory 2 89.674 907.44 143.18 2.1383 0.13147 partner.status 1 2.201 819.97 140.62 0.1050 0.74767 fcategory:partner.status 2 175.489 993.25 147.25 4.1846 0.02257 * --- Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1> Anova(mod.2, type=3)Anova Table (Type III tests) Response: conformity Sum Sq Df F value Pr(>F) (Intercept) 984.14 1 46.9348 3.436e-08 *** fcategory 89.67 2 2.1383 0.13147 partner.status 2.20 1 0.1050 0.74767 fcategory:partner.status 175.49 2 4.1846 0.02257 * Residuals 817.76 39 --- Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1 ---------------- snip ------------ I hope that this helps, John ----------------------------------------------- John Fox Senator McMaster Professor of Social Statistics Department of Sociology McMaster University Hamilton, Ontario, Canada> -----Original Message----- > From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] > On Behalf Of C. Weiner > Sent: Sunday, October 07, 2012 7:37 PM > To: r-help at r-project.org > Subject: [R] Why do I get different results for type III anova using the > drop1 or Anova command? > > Dear experts, > > I just noticed that I get different results conducting type III anova > using drop1 or the Anova command from the car package. I suppose I made > a mistake and hope you can offer me some help. I have no idea where I > got wrong and would be very grateful for explaination as R is new > terrain for me. > If I run the commands in line, they produce the same results. But if I > run them in seperate windows, they produce different results. > > What I did in easy words: > Opened R prompt, ran parts 0 and 1-> got result 1 > Opened new R prompt, ran parts 0 and 2 ->got result 2, ran part 1 > following part 2, got result 2a > > Why are the results different when they are run seperately and why do > the commands produce the same results when running one after the other? > Which results can I rely on? I based my script on: > http://mcfromnz.wordpress.com/2011/03/0 ... explained/ > <http://mcfromnz.wordpress.com/2011/03/02/anova-type-iiiiii-ss- > explained/> > but I surely made mistakes. > > Thanks for help. > Kind regards, > Christiane > > > > #0 > library(car) > P<-read.table("rhoPgegenrhoAgewichtet-mit Befruchtungssystem0912.txt", > sep="\t",dec=",", header=T,encoding="latin1") > summary(P) > str(P) > > #1 > modD3 <- lm(P$rhoPug ~ P$rhoAg*P$dprime*P$Befruchtungssystem, data=P) > > Anova(modD3,contrasts=list(rhoAg=contr.sum, dprime=contr.poly), > type="III") ###liefert dieses Ergebnis, wenn der Befehl als erster l?uft > #Response: P$rhoPug > # Sum Sq Df F value Pr(>F) > #(Intercept) 0.07719 1 4.5075 0.03530 * > #P$rhoAg 0.00687 1 0.4014 0.52730 > #P$dprime 0.00176 1 0.1029 0.74882 > #P$Befruchtungssystem 0.00250 1 0.1461 0.70280 > #P$rhoAg:P$dprime 0.00931 1 0.5439 0.46193 > #P$rhoAg:P$Befruchtungssystem 0.00001 1 0.0004 0.98343 > #P$dprime:P$Befruchtungssystem 0.00108 1 0.0630 0.80214 > #P$rhoAg:P$dprime:P$Befruchtungssystem 0.00206 1 0.1205 0.72893 > #Residuals 2.70584 158 > > > #2 > options(contrasts = c("contr.sum","contr.poly")) #liefert ein anderes > Ergebnis als der Befehl oben, wenn der obere Befehl NICHT gelaufen ist > modD3 <- lm(P$rhoPug ~ P$rhoAg*P$dprime*P$Befruchtungssystem, data=P) > drop1(modD3,.~.,test="F") # type III SS and F Test > P$rhoPug ~ P$rhoAg * P$dprime * P$Befruchtungssystem > Df Sum of Sq RSS AIC F value Pr(F) > <none> 2.7058 -667.35 > P$rhoAg 1 0.0131973 2.7190 -668.54 0.7706 0.3814 !! > P$dprime 1 0.0002240 2.7061 -669.34 0.0131 0.9091 !! > P$Befruchtungssystem 1 0.0025022 2.7083 -669.20 0.1461 0.7028 !! > P$rhoAg:P$dprime 1 0.0283677 2.7342 -667.62 1.6565 0.2000 !! > P$rhoAg:P$Befruchtungssystem 1 0.0000074 2.7058 -669.35 0.0004 0.9834 > P$dprime:P$Befruchtungssystem 1 0.0010790 2.7069 -669.29 0.0630 0.8021 > P$rhoAg:P$dprime:P$Befruchtungssystem 1 0.0020641 2.7079 -669.22 0.1205 > 0.7289 > > > #2a (command is the same as in 1, but results are different if the > command is run after the one above) > Anova(modD3,contrasts=list(Befruchtungssystem="contr.sum", > dprime="contr.poly"), type="III") #liefert in nachstehend nach dem > 2.Befehl dieses Ergebnis > #Response: P$rhoPug #also jetzt dasselbe wie drop1, aber nicht dasselbe, > das > #der exakt gleich Befehl alleine geliefert h?tte. Was > #stimmt den nun? Wo liegt mein Fehler? > # Sum Sq Df F value Pr(>F) > #(Intercept) 0.06877 1 4.0158 0.04679 * > #P$rhoAg 0.01320 1 0.7706 0.38136 > #P$dprime 0.00022 1 0.0131 0.90909 > #P$Befruchtungssystem 0.00250 1 0.1461 0.70280 > #P$rhoAg:P$dprime 0.02837 1 1.6565 0.19997 > #P$rhoAg:P$Befruchtungssystem 0.00001 1 0.0004 0.98343 > #P$dprime:P$Befruchtungssystem 0.00108 1 0.0630 0.80214 > #P$rhoAg:P$dprime:P$Befruchtungssystem 0.00206 1 0.1205 0.72893 > #Residuals 2.70584 158 > > ______________________________________________ > 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.