 Usage:  myfit = glmmultlo(x,y{,opt})  

 

 Input:



  x                      n x r matrix, the predictor variables. 

                         Individual-specific variables form single 

                         columns of x. Alternative-specific variables 

                         must be evaluated for each alternative and 

                         are to be stored in blocks of m subsequent 

                         columns. The optional parameters opt.indiv 

                         and opt.alter define which columns belong 

                         to which group. Without these optional 

                         parameters, all columns of x are interpreted 

                         as individual-specific. 

                         

  y                      n x 1 vector or n x m matrix. If y is vector 

                         (numeric or string), the different 

                         realizations are considered to be the 

                         alternatives. If y is matrix, it should 

                         contain 0/1 dummies, with 1 in the j-th column 

                         indicating that alternative j (in 1 ... m) has 

                         been chosen. 

                         

  opt                    optional, a list with optional input. The macro 

                         "glmopt" can be used to set up this parameter. 

                         The order of the list elements is not important. 

                         Parameters which are not given are replaced by 

                         defaults (see below). 

                         

  opt.shf                integer, if exists and =1, some output is produced 

                         which indicates how the iteration is going on. 

                         

  opt.miter              integer, maximal number of iterations. The default 

                         is 10. 

                         

  opt.cnv                scalar, convergence criterion. The default is 0.0001. 

                         

  opt.indiv              indices of columns of x (=variables) which are 

                         individual-specific, i.e. which are constant 

                         over alternatives. If neither opt.indiv and 

                         opt.alter are given, the model is estimated as 

                         if opt.indiv=1:r. Otherwise, the default is 

                         empty. 

                         

  opt.alter              indices of columns of x (=variables) which are 

                         alternative-specific, i.e. which vary over 

                         the alternatives (and eventually individuals). 

                         x[,opt.alter] should have a multiple of m 

                         columns, i.e. contain subsequent realizations 

                         for each of the alternatives. The default is empty. 

                         

 Output:



  myfit                  list with the components b, bv, and stat: 

                         

  b.indiv                p x m vector, estimated coefficients 

                         corresponding to individual-specific variables. 

                         The first column is zero. 

                         

  b.alter                q x 1 vector, estimated coefficients corresponding 

                         to alternative-specific variables. 

                         

  bv.indiv               (p*m) x (p*m) matrix, estimated covariance 

                         for b.indiv, matrix of p x p blocks. 

                         

  bv.alter               q x q matrix, estimated covariance for b.alter. 

                         

  bv.mixed               (p*m) x q matrix, estimated mixed covariances 

                         between b.alter and b.indiv, vector of p x q 

                         blocks. 

                         

  mu                     n x m vector, estimated response mu=P(y=j) 

                         corresponding to alternatives 1 ... m of y. 

                         

  stat                   list with the following statistics: 

                         

  stat.serror            standard errors, list containing components 

                         indiv and/or alter, respectively. 

                         

  stat.tvalues           t-values, list containing components 

                         indiv and/or alter, respectively. 

                         

  stat.pvalues           p-values, list containing components 

                         indiv and/or alter, respectively. 

                         

  stat.deviance          deviance, 

                         

  stat.loglik            log-likelihood, 

                         

  stat.r2                (pseudo) R^2. 

                         

  stat.adr2              adjusted (pseudo) R^2. 

                         

  stat.it                scalar, number of iterations needed 

                         

  stat.ret               scalar, return code: 

                         0 o.k., 

                         1 maximal number of iterations reached, 

                         -1 missing values have been encountered. 

                         

--------------------------------------------------------------

(C) MD*TECH Method and Data Technologies, 21.9.2000

