Hi,
I am new to clustering and was wondering why pvclust using "maximum"
as distance measure nearly always results in p-values above 95%.
I wrote an example programme which demonstrates this effect. I
uploaded a PDF showing the results
Here is the code which produces the PDF file:
-------------------------------------------------------------------------------------
s <- matrix(runif(1600,0,1000), nrow=20)
a.res1 <- pvclust(t(s), method.hclust="complete",
method.dist="euclidian",
nboot=500)
a.res2 <- pvclust(t(s), method.hclust="complete",
method.dist="maximum",
nboot=500)
a.res3 <- pvclust(t(s), method.hclust="complete",
method.dist="canberra",
nboot=500)
a.res4 <- pvclust(t(s), method.hclust="ward",
method.dist="euclidian",
nboot=500)
a.res5 <- pvclust(t(s), method.hclust="ward",
method.dist="maximum",
nboot=500)
a.res6 <- pvclust(t(s), method.hclust="ward",
method.dist="canberra",
nboot=500)
a.res7 <- pvclust(t(s), method.hclust="average",
method.dist="euclidian",
nboot=500)
a.res8 <- pvclust(t(s), method.hclust="average",
method.dist="maximum",
nboot=500)
a.res9 <- pvclust(t(s), method.hclust="average",
method.dist="canberra",
nboot=500)
s <- matrix(runif(1600,0,2), nrow=20)
b.res1 <- pvclust(t(s), method.hclust="complete",
method.dist="euclidian",
nboot=500)
b.res2 <- pvclust(t(s), method.hclust="complete",
method.dist="maximum",
nboot=500)
b.res3 <- pvclust(t(s), method.hclust="complete",
method.dist="canberra",
nboot=500)
b.res4 <- pvclust(t(s), method.hclust="ward",
method.dist="euclidian",
nboot=500)
b.res5 <- pvclust(t(s), method.hclust="ward",
method.dist="maximum",
nboot=500)
b.res6 <- pvclust(t(s), method.hclust="ward",
method.dist="canberra",
nboot=500)
b.res7 <- pvclust(t(s), method.hclust="average",
method.dist="euclidian",
nboot=500)
b.res8 <- pvclust(t(s), method.hclust="average",
method.dist="maximum",
nboot=500)
b.res9 <- pvclust(t(s), method.hclust="average",
method.dist="canberra",
nboot=500)
pdf("YOUR PATH/Res.pdf", width=10, height=15)
par(mfrow=c(6,3))
plot(a.res1)
plot(a.res2)
plot(a.res3)
plot(a.res4)
plot(a.res5)
plot(a.res6)
plot(a.res7)
plot(a.res8)
plot(a.res9)
plot(b.res1)
plot(b.res2)
plot(b.res3)
plot(b.res4)
plot(b.res5)
plot(b.res6)
plot(b.res7)
plot(b.res8)
plot(b.res9)
dev.off()
--------------------------------------------------------------
Why is that???
Cheers syrvn
http://r.789695.n4.nabble.com/file/n2295594/Res.pdf Res.pdf
--
View this message in context:
http://r.789695.n4.nabble.com/p-values-pvclust-maximum-distance-measure-tp2295594p2295594.html
Sent from the R help mailing list archive at Nabble.com.
Hi,
I am new to clustering and was wondering why pvclust using "maximum"
as distance measure nearly always results in p-values above 95%.
I wrote an example programme which demonstrates this effect. I
uploaded a PDF showing the results
Here is the code which produces the PDF file:
-------------------------------------------------------------------------------------
s <- matrix(runif(1600,0,1000), nrow=20)
a.res1 <- pvclust(t(s), method.hclust="complete",
method.dist="euclidian",
nboot=500)
a.res2 <- pvclust(t(s), method.hclust="complete",
method.dist="maximum",
nboot=500)
a.res3 <- pvclust(t(s), method.hclust="complete",
method.dist="canberra",
nboot=500)
a.res4 <- pvclust(t(s), method.hclust="ward",
method.dist="euclidian",
nboot=500)
a.res5 <- pvclust(t(s), method.hclust="ward",
method.dist="maximum",
nboot=500)
a.res6 <- pvclust(t(s), method.hclust="ward",
method.dist="canberra",
nboot=500)
a.res7 <- pvclust(t(s), method.hclust="average",
method.dist="euclidian",
nboot=500)
a.res8 <- pvclust(t(s), method.hclust="average",
method.dist="maximum",
nboot=500)
a.res9 <- pvclust(t(s), method.hclust="average",
method.dist="canberra",
nboot=500)
s <- matrix(runif(1600,0,2), nrow=20)
b.res1 <- pvclust(t(s), method.hclust="complete",
method.dist="euclidian",
nboot=500)
b.res2 <- pvclust(t(s), method.hclust="complete",
method.dist="maximum",
nboot=500)
b.res3 <- pvclust(t(s), method.hclust="complete",
method.dist="canberra",
nboot=500)
b.res4 <- pvclust(t(s), method.hclust="ward",
method.dist="euclidian",
nboot=500)
b.res5 <- pvclust(t(s), method.hclust="ward",
method.dist="maximum",
nboot=500)
b.res6 <- pvclust(t(s), method.hclust="ward",
method.dist="canberra",
nboot=500)
b.res7 <- pvclust(t(s), method.hclust="average",
method.dist="euclidian",
nboot=500)
b.res8 <- pvclust(t(s), method.hclust="average",
method.dist="maximum",
nboot=500)
b.res9 <- pvclust(t(s), method.hclust="average",
method.dist="canberra",
nboot=500)
pdf("YOUR PATH/Res.pdf", width=10, height=15)
par(mfrow=c(6,3))
plot(a.res1)
plot(a.res2)
plot(a.res3)
plot(a.res4)
plot(a.res5)
plot(a.res6)
plot(a.res7)
plot(a.res8)
plot(a.res9)
plot(b.res1)
plot(b.res2)
plot(b.res3)
plot(b.res4)
plot(b.res5)
plot(b.res6)
plot(b.res7)
plot(b.res8)
plot(b.res9)
dev.off()
--------------------------------------------------------------
Why is that???
Cheers syrvn
http://r.789695.n4.nabble.com/file/n2295593/Res.pdf Res.pdf
--
View this message in context:
http://r.789695.n4.nabble.com/p-values-pvclust-maximum-distance-measure-tp2295593p2295593.html
Sent from the R help mailing list archive at Nabble.com.