markovdata {depmix} R Documentation

## Specifying Markov data objects

### Description

Markovdata creates an object of class md, to be used by fitdmm.

### Usage


markovdata(dat, itemtypes, nitems = length(itemtypes), ntimes =
length(as.matrix(dat))/nitems, replicates = rep(1,
length(ntimes)), inames = NULL, dname = NULL, xm =
NA)

## S3 method for class 'md'
summary(object, ...)
## S3 method for class 'md'
plot(x, nitems = 1:(min(5, dim(x)[2])),
nind = 1:(min(5,length(attributes(x)\$ntimes))),...)
## S3 method for class 'md'
print(x, ...)

dname(object)
ntimes(object)
itemtypes(object)
replicates(object)

ncov(object)
inames(object)
nitems(object)
ind(object)



### Arguments

 dat An R object to be coerced to markovdata, a data frame or matrix. itemtypes A vector providing the types of measurement with possible values ‘continuous’, ‘categorical’, and ‘covariate’. This is mainly only used to rearrange the data when there are covariates in such a way that the covariate is in the last column. Only one covariate is supported in estimation of models. ntimes The number of repeated measurements, ie the length of the time series (this may be a vector containing the lengths of independent realiazations). It defaults the number of rows of the data frame or data matrix. replicates Using this argument case weights can be provided. This is particularly usefull in eg latent class analysis with categorical variables when there usually are huge numbers of replicates, ie identical response patterns. depmix computes the raw data log likelihood for each case separately. Thus, when there are many replicates of a case a lot of computation time is saved by specifying case weights instead of providing the full data set. inames The names of items. These default to the column names of matrices or dataframes. dname The name of the dataset, used in summary, print and plot functions. xm xm is the missing data code. It can be any value but zero. Missing data are recoded into NA. object,x An object of class md. ... Further arguments passed on to plot and summary. nitems,nind In the plot function, these arguments control which data are to be plotted, ie nitems indicates a range of items, and nind a range of realizations, respectively.

### Details

The function markovdata coerces a given data frame or matrix to be an object of class md such that it can be used in fitdmm. The md object has its own summary, print and plot methods.

The functions dname, itemtypes, ntimes, and replicates retrieve the respective attributes with these names; similarly ncov, nitems, inames, and ind retrieve the number of covariates, the number of items (the number of columns of the data), the column names and the number of independent realizations respectively.

### Value

An md-object is a matrix of dimensions sum(ntimes) by nitems, containing the measured variables and covariates rearranged such that the covariate appears in the last column. The column names are inames and the matrix has three further attributes:

 dname The name of the data set. itemtypes See above. ntimes See above. This will be a vector computed as ntimes=rep(ntimes,nreal). replicates The number of replications of each case, used as weigths in computing the log likelihood.

### Author(s)

Ingmar Visser i.visser@uva.nl

dmm, depmix

### Examples


x=rnorm(100,10,2)
y=ifelse(runif(100)<0.5,0,1)
z=matrix(c(x,y),100,2)
md=markovdata(z,itemtypes=c("cont","cat"))
summary(md)

data(speed)
summary(speed)
plot(speed,nind=2)

# split the data into three data sets
# (to perform multi group analysis)
r1=markovdata(dat=speed[1:168,],item=itemtypes(speed))
r2=markovdata(dat=speed[169:302,],item=itemtypes(speed))
r3=markovdata(dat=speed[303:439,],item=itemtypes(speed))
summary(r2)



[Package depmix version 0.9.16 Index]