bda (version 1.1.1-7)

binem: An E-M algorithm to fit normal mixture based on binned data.

Description

To fit a normal mixture based on data that are already binned.

Usage

binem(x,f,m=2,mu=NULL,na.rm=TRUE,just="center",binned=FALSE,
  scale=1.0, iter=100, modsel="AIC")

Arguments

x

vector of binned data; or can be a data frame of the binned data.

f

vector of frequencies if the data is not binned.

m

number of mixture components, default 2.

mu

A vector to sepcify the center of different component. If 'mu' is specified, 'm' will be ignored.

na.rm

Option to deal with missing value. Default: TRUE -- to remove missing values.

just

Justify the center of the bins for data that are rounded up ('right'), rounded down ('left'), or rounded to the nearest integers ('center').

binned

whether x is binned data or not. Default: FALSE.

scale

A positive value showing the scale of the measures in 'x'. Default 1 (data are rounded to the nearest integers). If the data are rounded to 100mg, scale=100.

iter

Iteration number in selecting initial values. Default 100.

modsel

Specify model selection criteria. Options: 'AIC', 'AICc' or 'BIC'.

See Also

dmixnorm, qmixnorm,rmixnorm.

Examples

x = c(rnorm(100,34.5,1.5), rnorm(200,40,1.39))
y = round(x,0)
(out = binem(y))
p = out$Para$Prop
mu = out$Para$Mean
s = out$Para$Std.Dev
x0 = seq(min(y),max(y),length=100)
plot(dmixnorm(x0,p,mu,s)~x0,type='l')
(out2 = smkde(y))
lines(out2,col=2)
lines(density(y),col=3)
lines(density(x),col=4)
p0 = c(1/3,2/3)
mu0 = c(34.5,40)
s0 = c(1.5,1.39)
lines(dmixnorm(x0,p0,mu0,s0)~x0,lwd=2,col=2)