bivariate.EM: EM-algorithm for bivariate normally distributed data

In CAMAN: Finite Mixture Models and Meta-Analysis Tools - Based on C.A.MAN

EM-algorithm for bivariate normally distributed data

Description

Function

Usage

bivariate.EM(obs1, obs2, type, data = NULL,
             var1, var2, 
             corr, lambda1, lambda2, p, 
             numiter=5000,acc=1.e-7,class)

Arguments

obs1	the first column of the observations

obs2	the second column of the observations

type	kind of data

data	an optional data frame. If not NULL, obs1, obs2, var1, var2 and corr will be looked for in data

var1	Variance of the first column of the observations(except meta-analysis)

var2	Variance of the second column of the observations (except meta-analysis)

corr	correlation coefficient

lambda1

Means of the first column of the observations

lambda2

Means of the second column of the observations

p	Mixing weight

numiter

parameter to control the maximal number of iterations in the EM loops. Default is 5000.

acc	convergence criterion. Default is 1.e-7

class

classification of studies?

Examples

## Not run: 
# 1.EM and classification for bivariate data with starting values
data(rs12363681)
lambda1<-c(1540.97, 837.12, 945.40, 1053.69)
lambda2<-c(906.66, 1371.81 ,1106.01,973.11)
p<-c(0.05,0.15,0.6,0.2)
test<-bivariate.EM(obs1=x, obs2=y, type="bi", lambda1=lambda1,lambda2=lambda2,
                   p=p,data=rs12363681,class="TRUE")
# scatter plot with ellipse
plot(test, ellipse=TRUE)
# scatter plot without ellipse
plot(test, ellipse=FALSE)

## End(Not run)
# 2. EM-algorithm for a diagnostic meta-analysis with bivariate 
#    normally distributed data and study specific fixed variances
data(CT)
p2<-c(0.4,0.6)
lamlog12<-c(2.93,3.22)
lamlog22<-c(2.5,1.5)
ct.m1 <- bivariate.EM(obs1=logitTPR, obs2=logitTNR, 
                      var1=varlogitTPR, var2=varlogitTNR,
                      type="meta", lambda1=lamlog12, lambda2=lamlog22,
                      p=p2,data=CT,class="TRUE")

CAMAN documentation built on Sept. 22, 2023, 5:12 p.m.