;config_template_linearize.scm
;lines starting with ; are comments
;Most of the options in this file are optional, but this file is
;a good check-list. Not all scm options
;are listed here, see the userguide for a complete list.
;;nonlinear model without any of the covariate combinations in test_relations.
;;Other covariates may be included
;;There is no file run34.mod in the extra material
model=run34.mod
;;it is helpful, but not necessary, to set the name of the run directory
directory=scm_run34_2
;---------------------------
;;options for linearization
linearize=1 ;must be set
;;foce=1 requires that nonlinear input model has METHOD=1. If nonlinear model
;;is run with FO set foce=0
foce=1
;;set update_derivatives=1 to increase accuracy by recalculating
;;derivatives after each covariate addition
update_derivatives=0
;;Set option lst_file if you want to use estimates from a lst-file
;;as initial estimates when running the nonlinear model to obtain
;;the derivatives needed as input to the linear model. Using
;;this option can save time.
lst_file=run34.lst
;;If the derivatives data has been generated previously, it is possible to
;;save time by giving this data file as input to the scm. Then there is
;;no need to run the nonlinear model before starting the forward search with
;;the linearized model. The user must check carefully that the derivatives
;;data file matches what is needed for this particular run, there is
;;no checking by the scm program.
;derivatives_data=run34_1_derivatives_covariates.dta
;;option epsilon is set by default, meaning that scm linearizes with respect
;;to epsilons in addition to with respect to etas.
;;try setting epsilon=0 if you run out of columns in $INPUT (NONMEM only
;;allows 50)
;epsilon=0
;;option error is only allowed if epsilon=0, then it is required.
;;This option tells scm the type of the error function.
;;The possible alternatives are add for additive, prop for proportional,
;;propadd for proportional plus additive or user for user-specified.
;error=add
;;if error=user then option error_code is required. Give exact NONMEM code
;;that spcifies Y using IPRED
;;If the code spans multiple lines all lines except the last must end
;;with \ (a backslash)
;error_code=Y=IPRED+IPRED*EPS(1)*EXP(ETA(4))+EPS(2)*EXP(ETA(5))
;;If any item in $INPUT is needed to run the model even for the base model
;;when no covariates are included it must be listed in do_not_drop.
;;Needed for all parameters used in IGNORE/ACCEPT statements and
;;for all parameters (except IPRED, ETA() and EPS()) in the user specified
;;error code, if option error_code is used.
;;Note: option do_not_drop has a slightly different functionality for
;;linearized and regular (linearize=0) scm. With linearize=1 do not list
;;parameters that are in $PK/$PRED
;;but not in IGNORE/ACCEPT or the error code.
;do_not_drop=AGE
;end options for linearization
;--------------------------------------------
;;search direction can be forward, backward or both
search_direction=forward
;;provided that the perl module Math::CDF is installed, any p-values can
;;be used. p_backward should be smaller than p_forward
p_forward=0.05
;p_backward=0.01
;;it is required to list the covariates to test
continuous_covariates=WGT,AGE
categorical_covariates=CVD1
;;if the covariates have missing data values, it is essential to
;;tell scm how missing values are coded
;missing_data_token=-99
;;if a parameter in test_relations is a logit transformation, it is
;;important to tell scm so that it can adjust the covariate
;;relation function properly
;logit=BIO,PHI
;;if parallel_states is set to 1, scm will test all possible relation forms
;;for a parameter-covariate pair simultaneously,
;;instead of trying them one at time, in the order set in valid_states
;parallel_states=1
;These general PsN options can be set in the configuration file.
;Most general PsN options must however
;be set on the command-line
nm_version=7 ;the linearize option only works together with NONMEM7
retries=1
threads=6
tweak_inits=1
picky=0
[test_relations]
CL=WGT,AGE
V=WGT,CVD1
;BIO=WGT
;PHI=WGT
;;valid_states (possibly in combination with [code]) tells scm which
;;parameterizations should be tested for the covariates
;;There are default meanings to numbers 1-5, but by adding a [code] section
;;new functions can be defined, and numbers can be set to mean different
;;parameterizations for different parameter-covariate pairs
;;The first valid state must always be 1
[valid_states]
continuous = 1,2,5
categorical = 1,2