Book: Measurement in medicine
Authors: HCW de Vet, CB Terwee, LB Mokkink, DL Knol
www.cambridge.org/9780521133852

* GET FILE chapter 8, assignment, database PRAFAB.

* select patients who are improved

USE ALL.
COMPUTE filter_$=(GRS9  <= 4).
VARIABLE LABEL filter_$ 'GRS9  <= 4 (FILTER)'.
VALUE LABELS filter_$  0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE .

* determine distribution chengeT0-T2 (Subtracting T2 values from T0 values leads to positive change scores for the improved patients) 

FREQUENCIES
  VARIABLES=ChangeT0_T2
  /ORDER=  ANALYSIS .


* select patients who are not importantly improved. note that the group 'slightly deteriorated' is also excluded.

FILTER OFF.
USE ALL.
EXECUTE .
USE ALL.
COMPUTE filter_$=(GRS9  > 4 AND GRS9 < 6).
VARIABLE LABEL filter_$ 'GRS9  > 4 AND GRS9 < 6 (FILTER)'.
VALUE LABELS filter_$  0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE .

* determine distribution over T0-T2 for the non importantly changed patients. 

FREQUENCIES
  VARIABLES=ChangeT0_T2
  /STATISTICS=MEAN STDDEV MIN MAX .


* determine the optimal ROC cut-off point of category GRS9 = 1-4 (n=345) versus GRS9 = 5 (n=139)
* we first create a new variable which we call ANCHOR, with
* ANCHOR = 1 when patients are importantly improved according to the anchor; 
* ANCHOR = 0 when patients are not importantly changed according to the anchor; and
* ANCHOR = -1 when patients are importantly deteriorated according to the anchor. 

IF (GRS9   <= 4) anchor = 1 .
EXECUTE .
IF (GRS9  =5) anchor = 0 .
EXECUTE .
IF (GRS9   >= 6) anchor = -1 .
EXECUTE .


FILTER OFF.
USE ALL.
EXECUTE .
USE ALL.
COMPUTE filter_$=(GRS9  <=  5).
VARIABLE LABEL filter_$ 'GRS9  <=  5 (FILTER)'.
VALUE LABELS filter_$  0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE .
ROC
  changeT0_T2  BY ANCHOR (1)
  /PLOT = CURVE
  /PRINT = SE COORDINATES
  /CRITERIA = CUTOFF(INCLUDE) TESTPOS(LARGE) DISTRIBUTION(FREE) CI(95)
  /MISSING = EXCLUDE .