Real World Health Care Data Analysis. Uwe Siebert
_meancont;
BY variable1 _strata_ _name_ _statn;
RETAIN stddiff;
IF UPCASE(_stat) = ‘MEAN’ THEN
DO;
stddiff = trt_1 - trt_0;
END;
ELSE IF UPCASE(_stat) = ‘STDDEV’ THEN
DO;
IF trt_1 ^= 0 THEN stddiff = stddiff / trt_1;
ELSE stddiff = .;
END;
IF LAST._name_;
KEEP variable1 _name_ stddiff _strata_; RUN;
PROC SORT DATA = _mean;
BY variable1 _strata_; RUN;
PROC TRANSPOSE DATA=_mean (DROP=_type_ _freq_) OUT=_mean_t PREFIX=trt_;
BY variable1 _strata_;
ID _cohort; RUN;
PROC SORT DATA = _mean_t;
BY variable1 _strata_; RUN;
DATA stddiff0_;
SET _mean_t;
trt_1 = FUZZ(trt_1);
var1 = trt_1 * ( 1 - trt_1);
IF var1 ^= 0 AND trt_0 NOT IN (0 1) THEN
stddiff = (trt_1 - trt_0) / SQRT(var1);
KEEP variable1 stddiff _strata_; RUN;
PROC SORT DATA = _meanclass;
BY variable1 _strata_;
RUN;
PROC TRANSPOSE DATA = _meanclass (DROP = _type_ _freq_) OUT =
_meanclass_t PREFIX = trt_;
BY variable1 _strata_;
ID _cohort; RUN;
PROC SORT DATA = _meanclass_t;
BY variable1 _strata_ _name_; RUN;
DATA stddiff2_ (RENAME = (_name_ = variable2));
SET _meanclass_t;
trt_1 = FUZZ(trt_1);
var1 = trt_1 * ( 1 - trt_1);
IF var1 ^= 0 AND trt_0 NOT IN (0 1) THEN
stddiff = (trt_1 - trt_0) / SQRT(var1);
KEEP variable1 _name_ stddiff _strata_; RUN;
DATA _stddiff;
SET _stddiff stddiff0_ (IN = in0) stddiff1_ (IN = in1) stddiff2_(IN = in2);
IF in0 THEN
DO;
vartype2 = ‘ ‘;
variable1 = UPCASE(variable1);
vartype1 = ‘C’;
END;
IF in1 THEN
DO;
variable2 = UPCASE(variable2);
_var = UPCASE(REVERSE(variable1));
IF NOT(variable2 =: REVERSE(SUBSTR(_var, INDEX(_var, ‘_’) + 1)));
vartype2 = ‘ ‘;
variable1 = UPCASE(variable1);
vartype1 = ‘C’;
END;
IF in2 THEN
DO;
variable2 = UPCASE(variable2);
_var = UPCASE(REVERSE(variable1));
IF NOT(variable2 =: REVERSE(SUBSTR(_var, INDEX(_var, ‘_’) + 1)));
vartype2 = ‘C’;
variable1 = UPCASE(variable1);
vartype1 = ‘C’;
END;
KEEP variable1 variable2 stddiff _strata_ vartype1 vartype2; RUN;
%endmac:
%MEND _ps_stddiff_apmb;
Program 4.9: Automatic Propensity Score Estimation Model
*************************************************************************
* Macro Name: PS_CALC_APMB *
* Parameters: *
* INDATA Name of the input dataset containing propensity scores *
* OUTDATA Name of the output dataset [&indata] *
* COHORT Name of variable containing the cohort/treatment variable*
* TREATED A string with the value of &COHORT denoting treated *
* patients
* CONTVARS List of continuous covariates to include in this table *
* CLASSVARS List of categorical covariates to include in this table *
* PS Name of variable to contain the propensity scores *
* CRIT The criteria to be used to consider an interaction term *
* balanced *
* MAXITER Maximum number of iterations allowed [800] *
* NSTRATA Number of strata [5] *
* IMBAL_STRATA_CRIT The criteria to be used to consider a strata *
* balanced *
* IMBAL_NSTRATA_CRIT Minimum number of imbalanced strata for a term *
* to be considered imbalanced. *
* ENTRY_NSTRATA_CRIT Minimum number of imbalanced strata for an *
* interaction term to be considered for entry into the model. *
* ALWAYS_INT Interaction term that are to always be included into the *
* model *
* DEBUG To reduce the amount of information written to the SAS log, *
* this parameter is set to NO by default [NO] | YES. *
************************************************************************;
%MACRO ps_calc_apmb (
indata = ,
outdata = ,
cohort = _cohort,
contvars = ,
classvars = ,
ps = ps,
treated = _treated,
n_mp_min = 100,
maxiter = 800,
imbal_strata_crit = 0.25,
imbal_nstrata_crit = 2,
entry_nstrata_crit = 2,
nstrata = 5,
debug = NO,
always_int =
);
%PUT Now executing macro %UPCASE(&sysmacroname);
%_ps_util;
%LET _notes = %SYSFUNC(GETOPTION(NOTES));
%LET _mprint = %SYSFUNC(GETOPTION(MPRINT));
%LET _mlogic = %SYSFUNC(GETOPTION(MLOGIC));
%LET _symbolgen = %SYSFUNC(GETOPTION(SYMBOLGEN));
ODS HTML CLOSE;
ODS LISTING CLOSE;
%* Checking parameter specifications;
%IF %BQUOTE(&indata) = %THEN
%DO;