pind = (1:rows(inds))~inds cind = 2|1 http://www...

quantlet.org/mdstat/codes/mva/MVAppexample.html (1 of 3) [10/11/2004 9:58:27 AM]
Applied Multivariate Statistical Analysis
setmaskp (pind, cind)
show (d, 2, 1, pind)
i = 0
pmin = 1
pmax = 1
while (i<n)
i = i+1
p = normal(cols(x))
p = p/sqrt(sum(p^2))
xp = x*p
ind = ppftind(xp, h)
inds = inds|ind
pind = (1:rows(inds))~inds
Code:
cind = 0.*matrix(rows(pind))
cind[maxind(inds)] = 4
cind[minind(inds)] = 1
cind[1] = 2
setmaskp(pind, cind)
show (d, 2, 1, pind)
if (ind>imax)
imax = ind
xa = setmask(grdot(xp), "red")
xa[,2] = -0.01-0.08.*xa[,2]
fa = setmask(denxest(xp, h, "rqua", xg), "line", "red")
show (d, 1, 1, zd, xa, fa, xi, fi)
pmin = i
endif
if (ind<imin)
imin = ind
xi = setmask(grdot(xp), "blue")
xi[,2] = -0.11-0.08*xi[,2]
fi = setmask(denxest(xp, h, "rqua", xg), "line", "blue")
show (d, 1, 1, zd, xa, fa, xi, fi)
pmax = i
endif
endo
endp
library ("plot")
library ("smoother")
library ("pp")
x = read ("bank2")
x = transform(x, grc.prep.sphere)
randomize(0)
h = 2.62.*rows(x)^(-1/5)
n = 50
ppexample(x, h, n)
http://www.quantlet.org/mdstat/codes/mva/MVAppexample.html (2 of 3) [10/11/2004 9:58:27 AM]
Applied Multivariate Statistical Analysis
22.10.2003
http://www.quantlet.org/mdstat/codes/mva/MVAppexample.html (3 of 3) [10/11/2004 9:58:27 AM]
Applied Multivariate Statistical Analysis
MVAsirdata
Description: MVAsirdata generates a data set and applies the sliced inverse regression algorithm (SIR) for dimension reduction.
Download: MVAsirdata.xpl
randomize(1998)
library("metrics")
n=300
x=normal(n,3)
e=normal(n,1)
b2=#(1,-1,-1)
b1=#(1,1,1)
i=#(1,2,3)
y=x*b1+((x*b1)^3)+4*((x*b2)^2)+e
{f,g}=sir(x,y,-20)
f1=(x*b1+((x*b1)^3)+4*((x*b2)^2))
m1=(x*b1)~y
m2=(x*b2)~y
m1=sort(m1)
m2=sort(m2)
sg=sum(g)
g=g/sg
psi=#(g[1],g[1]+g[2],g[1]+g[2]+g[3])
di=createdisplay(2,2)
hh=createdisplay(1,1)
kk=createdisplay(1,1)
ig=i~g
Code:
setmaskp(ig,0,12,8)
p11=x*f[,1]~y
setmaskp(p11,1,2,2)
p12=x*f[,1]~x*f[,2]~y
setmaskp(p12,1,2,2)
p21=x*f[,2]~y
setmaskp(p21,1,2,2)
show(di,1,1,p11)
show(di,2,1,p21)
show(di,1,2,p12)
show(di,2,2,ig,i~psi)
show(hh,1,1,m1)
show(kk,1,1,m2)
setgopt(hh,1,1,"title","True index vs Response","xlabel","first
index","ylabel","response")
http://www.quantlet.org/mdstat/codes/mva/MVAsirdata.html (1 of 2) [10/11/2004 9:58:38 AM]
Applied Multivariate Statistical Analysis
setgopt(kk,1,1,"title","True index vs Response","xlabel","second
index","ylabel","response")
setgopt(di,1,1,"title","XBeta1 vs Response","xlabel","first
index","ylabel","response")
setgopt(di,2,1,"title","XBeta2 vs Response","xlabel","second
index","ylabel","response")
setgopt(di,1,2,"title","XBeta1 XBeta2 Response")
setgopt(di,2,2,"title","Scree Plot","xlabel","K","ylabel","Psi(k) Eigelvalues") 22.10.2003
http://www.quantlet.org/mdstat/codes/mva/MVAsirdata.html (2 of 2) [10/11/2004 9:58:38 AM]
Applied Multivariate Statistical Analysis
MVAsir2data
MVAsir2data generates a data set and applies the sliced inverse regression algorithm (SIR II) for dimension
Description: reduction.
Download: MVAsir2data.xpl
randomize(1998)
library("metrics")
n=300
x=normal(n,3)
e=normal(n,1)
b2=#(1,-1,-1)
b1=#(1,1,1)
i=#(1,2,3)
y=x*b1+((x*b1)^3)+4*((x*b2)^2)+e
{f,g}=sir2(x,y,-20)
f1=(x*b1+((x*b1)^3)+4*((x*b2)^2))
m1=(x*b1)~(x*b1+((x*b1)^3)+4*((x*b2)^2))
m2=(x*b2)~(x*b1+((x*b1)^3)+4*((x*b2)^2))
m1=(x*b1)~y
m2=(x*b2)~y
m1=sort(m1)
m2=sort(m2)
sg=sum(g)
g=g/sg
psi=#(g[1],g[1]+g[2],g[1]+g[2]+g[3])
di=createdisplay(2,2)
hh=createdisplay(1,1)
kk=createdisplay(1,1)
Code:
ig=i~g
setmaskp(ig,0,12,8)
p11=x*f[,1]~y
setmaskp(p11,1,2,2)
p12=x*f[,1]~x*f[,2]~y
setmaskp(p12,1,2,2)
p21=x*f[,2]~y
setmaskp(p21,1,2,2)
show(di,1,1,p11)
show(di,2,1,p21)
show(di,1,2,p12)
show(di,2,2,ig,i~psi)
show(hh,1,1,m1)
http://www.quantlet.org/mdstat/codes/mva/MVAsir2data.html (1 of 2) [10/11/2004 9:58:59 AM]
Applied Multivariate Statistical Analysis
show(kk,1,1,m2)
setgopt(hh,1,1,"title","True index vs Response","xlabel","first
index","ylabel","response")
setgopt(kk,1,1,"title","True index vs Response","xlabel","second
index","ylabel","response")
setgopt(di,1,1,"title","XBeta1 vs Response","xlabel","first
index","ylabel","response")
setgopt(di,2,1,"title","XBeta2 vs Response","xlabel","second
index","ylabel","response")
setgopt(di,1,2,"title","XBeta1 XBeta2 Response")
setgopt(di,2,2,"title","Scree Plot","xlabel","K","ylabel","Psi(k) Eigelvalues") 22.10.2003
http://www.quantlet.org/mdstat/codes/mva/MVAsir2data.html (2 of 2) [10/11/2004 9:58:59 AM]
Applied Multivariate Statistical Analysis
MVAppsib
Description: Projection pursuit for the Boston housing data
Download: MVAppsib.xpl
proc(indmax)=ppexample(x, h, n)
xg = grid(-4, 0.1, 81)
zd = setmask(xg~pdfn(xg), "line", "green")
p = normal(cols(x))
p = p/sqrt(sum(p^2))
xp = x*p
indmin = p
indmax = p
imin = ppsj1ind(xp)
imax = imin
xi = setmask(grdot(xp), "blue")
xi[,2] = -0.11-0.08*xi[,2]
fi = setmask(denxest(xp, h, "rqua", xg), "line", "blue")
xa = setmask(grdot(xp), "red")
xa[,2] = -0.1-0.08*xa[,2]
fa = setmask(denxest(xp, h, "rqua", xg), "line", "red")
d = createdisplay (2,1)
setfractions(d,2|1,1)
show (d, 1, 1, zd, xa, fa, xi, fi)
inds = 0|imax
pind = (1:rows(inds))~inds
cind = 2|1
setmaskp (pind, cind)
show (d, 2, 1, pind)