mcmcordfactanal.rd

使用R语言的马尔科夫链蒙特卡洛模拟（MCMC）源代码程序。

\name{MCMCordfactanal}
\alias{MCMCordfactanal}
\title{Markov Chain Monte Carlo for Ordinal Data Factor Analysis Model}
\description{
  This function generates a sample from the posterior distribution of an
  ordinal data factor analysis model. Normal priors are assumed on the factor
  loadings and factor scores while improper uniform priors are assumed
  on the cutpoints. The user supplies data and parameters for the prior
  distributions, and a sample from the posterior distribution is returned as
  an mcmc object, which can be subsequently analyzed with
  functions provided in the coda package.
}
  
\usage{
MCMCordfactanal(x, factors, lambda.constraints=list(),
                data=parent.frame(), burnin = 1000, mcmc = 20000,
                thin=1, tune=NA, verbose = 0, seed = NA,
                lambda.start = NA, l0=0, L0=0,
                store.lambda=TRUE, store.scores=FALSE,
                drop.constantvars=TRUE, ... )
 }

\arguments{
  \item{x}{Either a formula or a numeric matrix containing the
    manifest variables.}

  \item{factors}{The number of factors to be fitted.}
  
  \item{lambda.constraints}{List of lists specifying possible equality
    or simple inequality constraints on the factor loadings. A typical
    entry in the list has one of three forms: \code{varname=list(d,c)} which
    will constrain the dth loading for the variable named varname to
    be equal to c, \code{varname=list(d,"+")} which will constrain the dth
    loading for the variable named varname to be positive, and
    \code{varname=list(d, "-")} which will constrain the dth loading for the
    variable named varname to be negative. If x is a matrix without
    column names defaults names of ``V1", ``V2", ... , etc will be
    used. Note that, unlike \code{MCMCfactanal}, the
    \eqn{\Lambda}{Lambda} matrix used here has \code{factors}+1
    columns. The first column of \eqn{\Lambda}{Lambda} corresponds to
    negative item difficulty parameters and should generally not be
    constrained. 
    }
  
  \item{data}{A data frame.}
  
  \item{burnin}{The number of burn-in iterations for the sampler.}
  
  \item{mcmc}{The number of iterations for the sampler.}
  
  \item{thin}{The thinning interval used in the simulation.  The number of
    iterations must be divisible by this value.}

  \item{tune}{The tuning parameter for the Metropolis-Hastings
    sampling. Can be either a scalar or a \eqn{k}{k}-vector. Must be
    strictly positive.}
  
  \item{verbose}{A switch which determines whether or not the progress of
    the sampler is printed to the screen.  If \code{verbose} is greater
    than 0 the iteration number and
    the Metropolis-Hastings acceptance rate are printed to the screen
    every \code{verbose}th iteration.}
  
    \item{seed}{The seed for the random number generator.  If NA, the Mersenne
    Twister generator is used with default seed 12345; if an integer is 
    passed it is used to seed the Mersenne twister.  The user can also
    pass a list of length two to use the L'Ecuyer random number generator,
    which is suitable for parallel computation.  The first element of the
    list is the L'Ecuyer seed, which is a vector of length six or NA (if NA 
    a default seed of \code{rep(12345,6)} is used).  The second element of 
    list is a positive substream number. See the MCMCpack 
    specification for more details.}
  
  \item{lambda.start}{Starting values for the factor loading matrix
    Lambda. If \code{lambda.start} is set to a scalar the starting value for
    all unconstrained loadings will be set to that scalar. If
    \code{lambda.start} is a matrix of the same dimensions as Lambda then the
    \code{lambda.start} matrix is used as the starting values (except
    for equality-constrained elements). If \code{lambda.start} is set to
    \code{NA} (the default) then starting values for unconstrained
    elements in the first column of Lambda are based on the observed
    response pattern, the remaining unconstrained elements of Lambda are
    set to , and starting values for inequality constrained elements are
    set to either 1.0 or -1.0 depending on the nature of the constraints.}
  
  \item{l0}{The means of the independent Normal prior on the factor
    loadings. Can be either a scalar or a matrix with the same
    dimensions as \code{Lambda}.}

  \item{L0}{The precisions (inverse variances) of the independent Normal
    prior on the factor loadings. Can be either a scalar or a matrix with
    the same dimensions as \code{Lambda}.}

  \item{store.lambda}{A switch that determines whether or not to store
    the factor loadings for posterior analysis. By default, the factor
    loadings are all stored.}
  
  \item{store.scores}{A switch that determines whether or not to
    store the factor scores for posterior analysis. 
    \emph{NOTE: This takes an enormous amount of memory, so
      should only be used if the chain is thinned heavily, or for
      applications with a small number of observations}.  By default, the
    factor scores are not stored.}

  \item{drop.constantvars}{A switch that determines whether or not
        manifest variables that have no variation should be deleted
	before fitting the model. Default = TRUE.}

  \item{...}{further arguments to be passed}       
}

\value{
   An mcmc object that contains the posterior sample.  This 
   object can be summarized by functions provided by the coda package.
}

\details{The model takes the following form:

  Let \eqn{i=1,\ldots,N}{1=1,...,n} index observations and
  \eqn{j=1,\ldots,K}{j=1,...,K} index response variables within an
  observation. The typical observed
  variable \eqn{x_{ij}}{x_ij} is ordinal with a total of \eqn{C_j}{C_j}  
  categories. The distribution of \eqn{X}{X} is governed by a \eqn{N
    \times K}{N by K} matrix of latent variables \eqn{X^*}{Xstar} and a
  series of cutpoints \eqn{\gamma}{gamma}. \eqn{X^*}{Xstar} is assumed
  to be generated according to:
  
  \deqn{x^*_i = \Lambda \phi_i + \epsilon_i}{xstar_i = Lambda phi_i +
    epsilon_i}
  \deqn{\epsilon_i \sim \mathcal{N}(0,I)}{epsilon_i ~ N(0, I)}

  where \eqn{x^*_i}{xstar_i} is the \eqn{k}{k}-vector of latent variables
  specific to observation \eqn{i}{i}, \eqn{\Lambda}{Lambda} is the
  \eqn{k \times d}{k by d} matrix of factor loadings, and
  \eqn{\phi_i}{phi_i} is
  the \eqn{d}{d}-vector of latent factor scores. It is assumed that the
  first element of \eqn{\phi_i}{phi_i} is equal to 1 for all
  \eqn{i}{i}. 

  The probability that the \eqn{j}{j}th variable in observation
  \eqn{i}{i} takes the value \eqn{c}{c} is:

   \deqn{
     \pi_{ijc} = \Phi(\gamma_{jc} - \Lambda'_j\phi_i) -
     \Phi(\gamma_{j(c-1)} - \Lambda'_j\phi_i)
   }{
     pi_ijc = pnorm(gamma_jc - Lambda'_j phi_i) -
     pnorm(gamma_j(c-1) - Lambda'_j phi_i)
   }
  
  The implementation used here assumes independent conjugate priors for
  each element of \eqn{\Lambda}{Lambda} and each
  \eqn{\phi_i}{phi_i}. More specifically we assume:

  \deqn{\Lambda_{ij} \sim \mathcal{N}(l_{0_{ij}}, L_{0_{ij}}^{-1}),
    i=1,\ldots,k,  j=1,\ldots,d}{Lambda_ij ~ N(l0_ij,  L0_ij^-1),
    i=1,...,k, j=1,...,d} 

  \deqn{\phi_{i(2:d)} \sim \mathcal{N}(0, I),
    i=1,\dots,n}{phi_i(2:d) ~ N(0, I),
      i=1,...,n} 


    The standard two-parameter item response theory model with probit
    link is a special case of the model sketched above. 
    
  \code{MCMCordfactanal} simulates from the posterior distribution using
  a Metropolis-Hastings within Gibbs sampling algorithm. The algorithm
  employed is based on work by Cowles (1996).  Note that
  the first element of \eqn{\phi_i}{phi_i} is a 1. As a result, the
  first column of \eqn{\Lambda}{Lambda} can be interpretated as item
  difficulty parameters.  Further, the first
  element  \eqn{\gamma_1}{gamma_1} is normalized to zero, and thus not 
  returned in the mcmc object.
  The simulation proper is done in compiled C++ code to maximize
  efficiency.  Please consult the coda documentation for a comprehensive
  list of functions that can be used to analyze the posterior
  sample. 
  
  As is the case with all measurement models, make sure that you have plenty
  of free memory, especially when storing the scores.
  }

\references{
  Shawn Treier and Simon Jackman. 2003. ``Democracy as a Latent Variable." 
  Paper presented at the Midwest Political Science Association Annual Meeting.

  M. K. Cowles. 1996. ``Accelerating Monte Carlo Markov Chain Convergence for
  Cumulative-link Generalized Linear Models." \emph{Statistics and Computing.}
  6: 101-110.
     
  Valen E. Johnson and James H. Albert. 1999. ``Ordinal Data Modeling." 
  Springer: New York.
  
   Daniel Pemstein, Kevin M. Quinn, and Andrew D. Martin.  2007.  
   \emph{Scythe Statistical Library 1.0.} \url{http://scythe.wustl.edu}.
   
   Martyn Plummer, Nicky Best, Kate Cowles, and Karen Vines. 2002.
   \emph{Output Analysis and Diagnostics for MCMC (CODA)}.
   \url{http://www-fis.iarc.fr/coda/}.
}


\examples{
   \dontrun{
   data(painters)
   new.painters <- painters[,1:4]
   cuts <- apply(new.painters, 2, quantile, c(.25, .50, .75))
   for (i in 1:4){
      new.painters[new.painters[,i]<cuts[1,i],i] <- 100
     new.painters[new.painters[,i]<cuts[2,i],i] <- 200
     new.painters[new.painters[,i]<cuts[3,i],i] <- 300
     new.painters[new.painters[,i]<100,i] <- 400
   }

   posterior <- MCMCordfactanal(~Composition+Drawing+Colour+Expression,
                        data=new.painters, factors=1,
                        lambda.constraints=list(Drawing=list(2,"+")),
                        burnin=5000, mcmc=500000, thin=200, verbose=500,
                        L0=0.5, store.lambda=TRUE,
                        store.scores=TRUE, tune=1.2)
   plot(posterior)
   summary(posterior)
   }
}

\keyword{models}

\seealso{\code{\link[coda]{plot.mcmc}}, \code{\link[coda]{summary.mcmc}},
  \code{\link[mva]{factanal}}, \code{\link[MCMCpack]{MCMCfactanal}},
  \code{\link[MCMCpack]{MCMCirt1d}}, \code{\link[MCMCpack]{MCMCirtKd}}}