SPR-PSplines: Semiparametric Approaches to Model the Prevalence and Force...

In TeaKov/serostat: Modeling Infectious Disease Parameters Based on Serological and Social Contact Data R Package

Description

Univariate smoother using P-splines, by Paul Eilers & Brian Marx (1995). For the smooth splines, thin plate splines, and adaptive smoothing splines approaches, go to help("SemiParametricRegression").

Usage

pspline.fit(response, x.var, ps.intervals, wts, degree, order,
	link, family, m.binomial, r.gamma, lambda, x.predicted, ridge.adj)

Arguments

response	Response variable
x.var	Explanatory variable on abcissae
ps.intervals	Number of intervals for B-splines. Default=8.
wts	Vector of weights; default is vector of ones.
degree	Degree of B-splines. Default=3.
order	Order of difference penalty. Default=3.
link	Link function (identity, log, sqrt, logit, probit, cloglog, loglog, recipical).
family	What kind of distribution (family=gaussian, binomial, poisson, gamma)
m.binomial	Vector of binomial trials. Default is 1 vector.
r.gamma	Vector of gamma shape parameters. Default is 1 vector.
lambda	Smoothness regularizing parameter ( >= 0). Default=0.
x.predicted	A list of x variables for prediction and twice stderr limits.
ridge.adj	Default=0.0001

Examples

# Load Belgian B19 data
data("VZV_B19_BE_0103")
subset<-(VZV_B19_BE_0103$age>0.5)&(VZV_B19_BE_0103$age<76)&(!is.na(VZV_B19_BE_0103$age))&
  !is.na(VZV_B19_BE_0103$parvores)
VZV_B19_BE_0103<-VZV_B19_BE_0103[subset,]
y<-VZV_B19_BE_0103$parvores[order(VZV_B19_BE_0103$age)]
a<-VZV_B19_BE_0103$age[order(VZV_B19_BE_0103$age)]
s<-VZV_B19_BE_0103$sex[order(VZV_B19_BE_0103$age)]

grid<-sort(unique(round(a)))
neg<-table(y,round(a))[1,]
pos<-table(y,round(a))[2,]
tot<-neg+pos

degree <- 3
order <- 2

# P-splines, with logit link-function
lopt <- 80
pspline.fit(response=y, x.var=a, ps.intervals=20, degree=degree, order=order,
  link="logit", family="binomial", lambda=lopt, x.predicted=a)

# P-splines, with cloglog link-function
lopt <- 200
pspline.fit(response=y, x.var=a, ps.intervals=20, degree=degree, order=order,
  link="cloglog", family="binomial", lambda=lopt, x.predicted=a)

TeaKov/serostat documentation built on May 21, 2019, 1:21 p.m.