

   OOrrddiinnaall RRaannddoomm EEffffeeccttss MMooddeellss wwiitthh DDrrooppoouuttss

        logitord(y, id, out.ccov=NULL, drop.ccov=NULL, tvcov=NULL,
             out.tvcov=!is.null(tvcov), drop.tvcov=!is.null(tvcov),
             pout, pdrop, prand.out, prand.drop,
             random.out.int=T, random.out.slope=!is.null(tvcov),
             random.drop.int=T, random.drop.slope=!is.null(tvcov),
             binom.mix=5, fcalls=900, eps=0.0001, print.level=0)

   AArrgguummeennttss::

          y: A vector of binary or ordinal responses with lev-
             els 1 to k and 0 indicating drop-out.

         id: Identification number for each individual.

   out.ccov: A vector, matrix, or model formula of time-con-
             stant covariates for the outcome regression, with
             variables having the same length as y.

   drop.ccov: A vector, matrix, or model formula of time-con-
             stant covariates for the drop-out regression, with
             variables having the same length as y.

      tvcov: One time-varying covariate vector.

   out.tvcov: Include the time-varying covariate in the outcome
             regression.

   drop.tvcov: Include the time-varying covariate in the drop-
             out regression.

       pout: Initial estimates of the outcome regression coef-
             ficients, with length equal to the number of lev-
             els of the response plus the number of covariates
             minus one.

      pdrop: Initial estimates of the drop-out regression coef-
             ficients, with length equal to one plus the number
             of covariates.

   prand.out: Optional initial estimates of the outcome random
             parameters.

   prand.drop: Optional initial estimates of the drop-out ran-
             dom parameters.

   random.out.int: If TRUE, the outcome intercept is random.

   random.out.slope: If TRUE, the slope of the time-varying
             covariate is random for the outcome regression
             (only possible if a time-varying covariate is sup-
             plied and if out.tvcov and random.out.int are
             TRUE).

   random.drop.int: If TRUE, the drop-out intercept is random.

   random.drop.slope: If TRUE, the slope of the time-varying
             covariate is random for the drop-out regression
             (only possible if a time-varying covariate is sup-
             plied and if drop.tvcov and random.drop.int are
             TRUE).

   binom.mix: The total in the binomial distribution used to
             approximate the normal mixing distribution.

     fcalls: Number of function calls allowed.

        eps: Convergence criterion.

   print.level: If 1, the iterations are printed out.

   DDeessccrriippttiioonn::

        `logitord' fits an longitudinal ordinal model in dis-
        crete time to outcomes and a logistic model to the
        probability of dropping out using a common random
        effect for the two.

   VVaalluuee::

        A list of class `logitord' is returned.

   AAuutthhoorr((ss))::

        T.R. Ten Have and J.K. Lindsey

   RReeffeerreenncceess::

        Ten Have, T, Kunselman, A.R., Pulkstenis, E.P. and Lan-
        dis, J.R.  (1998) Biometrics 54, 367-383, for the
        binary case.

   EExxaammpplleess::

        y <- trunc(runif(20,max=4))
        id <- gl(4,5)
        age <- rpois(20,20)
        times <- rep(1:5,4)
        logitord(y, id=id, out.ccov=~age, drop.ccov=age, pout=c(1,0,0),
             pdrop=c(1,0))
        logitord(y, id, tvcov=times, pout=c(1,0,0), pdrop=c(1,0))

