 Usage:  {b,bv,df,m,mg,it}  = gplmcore(code,x,t,y,h,wx,wt,wc,b0,m0,ctrl{,upb{,tg,m0g}})  
 
 Input:

  code                   text string, the short code for the model (e.g. 
                         "bilo" for logit or "noid" for ordinary PLM). 
                         
  x                      n x p matrix, the discrete predictor variables. 
                         
  t                      n x q matrix, the continuous predictor variables. 
                         Needs to be SORTED by the first column. 
                         
  y                      n x d vector, the response variables. 
                         
  h                      q x 1 vector, the bandwith. 
                         
  wx                     n x 1 vector or scalar, prior weights, e.g. the 
                         binomial index vector. 
                         
  wt                     n x 1 vector or scalar, weights for t (trimming 
                         factors). Is ignored, when scalar. 
                         
  wc                     n x 1 vector or scalar, weights for convergence 
                         criterion, w.r.t. m(t) only. Is ignored, when scalar. 
                         
  b0                     p x 1 vector, the initial coefficients. 
                         
  m0                     n x 1 vector or scalar, the initial values for the 
                         nonparametric part. Is ignored and can be set to 
                         scalar direct update for nonparametric function is 
                         possible (code="noid"). 
                         
  off                    n x 1 vector or scalar, offset. Is ignored when 0. 
                         
  ctrl                   7 x 1 vector or scalar, contains control parameters 
                         shf (default=0), 
                         miter (default=10), 
                         cnv (default=0.0001), 
                         fscor (default=0), 
                         pow (default=0, power for power link), 
                         nbk (default=1, parameter for negative binomial), 
                         meth (default=0, parameter for backfitting/profile). 
                         Alternatively, one can give here shf only. Set to 0 
                         to use the defaults. 
                         The parameters correspond to the optional parameters 
                         which can be given in gplminit. 
                         They are all ignored when not applicable. 
                         
  upb                    optional, scalar, 0 or 1 (default). If set to 
                         0, the parameter b is not updated in the 
                         iteration. 
                         
  tg                     optional, ng x 1 vector, a grid for continuous part. 
                         Needs to be SORTED by the first column. Is ignored, 
                         if set to NaN. 
                         
  m0g                    optional, ng x 1 vector or scalar, the initial values 
                         for the nonparametric part on the grid. Needs to be 
                         given if direct update for nonparametric function 
                         is not possible. Is ignored otherwise. Is ignored, 
                         if tg set to NaN. 
                         
 Output:

  b                      p x 1 vector, estimated coefficients 
                         
  bv                     p x p matrix, estimated covariance matrix for coeff. 
                         
  m                      n x 1 vector, estimated nonparametric part. 
                         
  df                     scalar, approximated degrees of freedom. 
                         
  mg                     ng x 1 vector, estimated nonparametric part on grid. 
                         Only available if tg was given. 
                         
  it                     integer, number of iterations needed. 
                         
  ret                    scalar, return code: 
                         0 o.k., 
                         1 maximal number of iterations reached 
                         (if applicable), 
                         -1 missing values have been encountered. 
                         
--------------------------------------------------------------
(C) MD*TECH Method and Data Technologies, 17.8.2000
