AD Model Builder/Runtime Options

From QERM Wiki
Jump to: navigation, search

The following is the runtime options for ADMB. This can also be accessed from the -? option.

AD Model Builder version 7.1.1 copyright (c) 1993 1998 2005 Otter Research Ltd
USAGE--kalman options
where an option consists of -option_tag followed by arguments if necessary
-ainp NAME change default ascii input parameter file name to NAME
-binp NAME change default binary input parameter file name to NAME
-est only do the parameter estimation
-ind NAME change default input data file name to NAME
-lmn N use limited memory quasi newton -- keep N steps
-lprof perform profile likelihood calculations
-prsave save the independent variables from the profile calculations
-maxph N increase the maximum phase number to N
-mcdiag use diagonal covariance matrix for mcmc with diagonal values 1
-mcmc [N] perform markov chain monte carlo with N simulations
-mcmult N multiplier N for mcmc default
-mcr resume previous mcmc
-mcrb N modify the covariance matrix to reduce extremely high correlation
-mcnoscale don’t rescale step size for mcmc depending on acceptance rate
-mcgrope N use probing strategy for mcmc with factor N
-mcseed N seed for random number generator for markov chain monte carlo
-mccale N rescale step size for first N evaluations
-mcsave N save the parameters for every N’th simulation
-mceval Go throught the saved mcmc values from a previous mcsave
-mcpin NAME Read the starting values for MCMC from the file NAME
-crit N set gradient magnitude convergence criterion to N
-iprint N print out function minimizer report every N iterations
-maxfn N set maximum number opf function eval’s to N
-rs if function minmimizer can’t make progress rescale and try again
-nox don’t show vector and gradient values in function minimizer screen report
-phase N start minimization in phase N
-simplex use simplex algorithm for minimization (new test version)
-sdonly do delta method for std dev estimates without redoing hessian
-ams N set arrmblsize to n (ARRAY_MEMBLOCK_SIZE)
-cbs N set CMPDIF_BUFFER_SIZE TO N
-mno N set the maximum number of independent variables to N
-gbs N set GRADSTACK_BUFFER_SIZE TO N
-mdl N set the maximum number of dvariables to N
Personal tools