| Library: | gam |
| See also: | intest intestpl gintestpl gamfit pcad |
| Quantlet: | gintest | |
| Description: | estimation of the univariate additive functions in a separable generalized additive model using Nad.Watson, local linear or local quadratic |
| Usage: | m = gintest(code,t,y,h,g,loc{,opt}) | |
| Input: | ||
| code | string, specifying the code function implemented codes: noid, bipro, bilo | |
| t | n x p matrix, the continuous predictor variables. | |
| y | n x q matrix , the observed response variables | |
| h | p x 1 or 1 x 1 matrix , chosen bandwidth for the directions of interest | |
| g | p x 1 or 1 x 1 matrix , chosen bandwidth for the directions not of interest | |
| loc | dummy , for loc=0 local constant (Nad. Wats.), for loc=1 local linear and for loc=2 local quadratic estimator will be used | |
| opt | optional, a list with optional input. The macro "gplmopt" can be used to set up this parameter. The order of the list elements is not important. | |
| opt.tg | ng x pg vector, a grid for continuous part. If tg is given, the nonparametric function will also be computed on this grid. | |
| opt.shf | integer, (show-how-far) if exists and =1, an output is produced which indicates how the iteration is going on (additive function / point of estimation / number of iteration). | |
| Output: | ||
| m | n(ng) x p(pg) x q matrix, containing the marginal integration estimators | |
library("gam")
randomize(1235)
n = 100
p = 2
t = uniform(n,p)*2-1
g1 = 2*t[,1]
g2 = t[,2]^2
g2 = g2 - mean(g2)
m = g1 + g2
y = cdfn(m) .> uniform(n) ; probit model
h = #(1.7, 1.5)
g = #(1.7, 1.5)
tg = grid(-0.8,0.1,19)
opt = gamopt("tg",tg~tg,"shf",1)
loc = 1
code = "bipro"
m = gintest(code,t,y,h,g,loc,opt)
d1 = tg[,1]~m[,1]
d2 = tg[,2]~m[,2]
setmaskp(d1,4,4,8)
setmaskp(d2,4,4,8)
bild = createdisplay(1,2)
show(bild,1,1,d1,t[,1]~g1)
show(bild,1,2,d2,t[,2]~g2)
the marginal integration estimator of the additive functions, see Linton and Haerdle (1996)