NOTE: Copyright (c) 1999-2001 by SAS Institute Inc., Cary, NC, USA.
NOTE: SAS (r) Proprietary Software Release 8.2 (TS2M0)
      Licensed to GEORGE WASHINGTON UNIVERSITY-CAMPUSWIDE-T/R, Site 0001180006.
NOTE: This session is executing on the WIN_ME  platform.
 
 
 
NOTE: SAS initialization used:
      real time           19.23 seconds
 
1    *** sh_b from old ST210H;
2
3    options formdlim='-';
4
5    filename sh_blst 'c:\sasjobs\stat210\jl2003\sh_blst.pdf';
6    ods printer pdf file=sh_blst;
NOTE: Writing ODS PRINTER output to DISK destination "c:\sasjobs\stat210\jl2003\sh_blst.pdf", printer "PDF".
7
8    %macro runloess (data=_last_,out=,x=,y=,f=0.5);
9    proc loess data=&data;
10   model &y = &x /smooth=&f;
11   ods output OutputStatistics=&out;
12   title2 "smoothed plot of &y versus &x with smooth &f";
13   proc plot; by SmoothingParameter;
14     plot DepVar*&x='*' Pred*&x='+' / overlay;
15   run;
16   %mend;
17
18   %MACRO rsquare;
19   data; set end=eof;
20   retain lab n1 nT;
21   if _n_ = 1 then do;
22     lab=1; n1=0; nt=0;
23     end;
24   if oy=1 then n1 = n1+1;
25   if oy ge 0 then do;
26      nt=nt+1;
27      lab = lab * ( (p**oy) * ((1-p)**(1-oy)) );
28   end;
29   if eof then do;
30     alpha = log( (n1/nt)/(1 - (n1/nt)) );
31     pa = exp(alpha) / (1 + exp(alpha));
32     la = (pa**n1) * ( ((1-pa)**(nt-n1)) );
33     nullchi = -2*log(la);
34     abchi = -2*log(lab);
35     modchi =  nullchi - abchi;
36     r2ent = modchi/nullchi;
37     r2chi = 1 - exp(-modchi/nt);
38     output;
39   end;
40   proc print; var n1 nt pa la lab nullchi abchi modchi r2ent r2chi;
41   title3 'cross validation r-square measures';
42
43   %MEND;
44   RUN;
45
46   DATA FITNESS;
47    INPUT AGE WEIGHT OXY RUNTIME RSTPULSE RUNPULSE MAXPULSE NONSMOKE;
48    N=1;
49    Y=0; IF OXY > 47 THEN Y = 1;    *****NONSMOKE=1=NONSMOKER;
50                                    *****         0=SMOKER;
51   CARDS;
 
NOTE: The data set WORK.FITNESS has 31 observations and 10 variables.
NOTE: DATA statement used:
      real time           0.65 seconds
 
 
83   ;
84
85   PROC LOGISTIC DESCENDING DATA = FITNESS OUTEST=BETAS;
86        MODEL Y= AGE WEIGHT RUNTIME RUNPULSE NONSMOKE
87     / clodds=both lackfit rsquare CORRB INFLUENCE IPLOTS;
88     OUTPUT OUT=DUM1 XBETA=XB STDXBETA=SDXB L=L P=P U=U;
89   TITLE2 'LOGISTIC REGRESSION WITH ALL VARIABLES';
90
 
NOTE: PROC LOGISTIC is modeling the probability that Y=1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
NOTE: There were 31 observations read from the data set WORK.FITNESS.
NOTE: The data set WORK.BETAS has 1 observations and 11 variables.
NOTE: The data set WORK.DUM1 has 31 observations and 16 variables.
NOTE: PROCEDURE LOGISTIC used:
      real time           3.19 seconds
 
 
91   PROC GENMOD DATA = FITNESS;
92     MODEL Y/N = AGE WEIGHT RUNTIME RUNPULSE NONSMOKE
93     / DIST=BIN LINK=LOGIT TYPE3;
94   TITLE2 'LOGISTIC REGRESSION THROUGH GENMOD';
95
 
NOTE: Algorithm converged.
NOTE: The scale parameter was held fixed.
NOTE: PROCEDURE GENMOD used:
      real time           1.04 seconds
 
 
96   PROC PRINT DATA=DUM1; VAR AGE WEIGHT RUNTIME RUNPULSE NONSMOKE
97     Y XB SDXB L P U;
98
 
NOTE: There were 31 observations read from the data set WORK.DUM1.
NOTE: PROCEDURE PRINT used:
      real time           0.04 seconds
 
 
99   DATA PARTIAL;
100    RETAIN B0 B1 B2 B3 B4 B5;
101    ARRAY B (5) B1-B5;
102    ARRAY X (5)
103                AGE WEIGHT RUNTIME RUNPULSE NONSMOKE;
104    ARRAY RES (5) RES1-RES5;
105    IF _N_=1 THEN SET BETAS
106       (RENAME = (INTERCEPt=B0
107        AGE=B1 WEIGHT=B2 RUNTIME=B3 RUNPULSE=B4 NONSMOKE=B5));
108        SET FITNESS;
109  ETA=B0;
110  DO J = 1 TO 5;
111    ETA = ETA + (X(j)*B(j));
112  END;
113  P = 1/(1+EXP(-ETA));
114  DO J = 1 TO 5;
115    RES(J) = B(J)*X(J) + (Y-P)/(P*(1-P));
116  END;
117
118  title2 'partial residual plots with lowess smother';
119  TITLE3 'Scatter-plots of partial residuals for each covariate';
120  %runloess (data=partial,out=smooth,x=age,y=res1,f=0.5);
 
NOTE: There were 1 observations read from the data set WORK.BETAS.
NOTE: There were 31 observations read from the data set WORK.FITNESS.
NOTE: The data set WORK.PARTIAL has 31 observations and 29 variables.
NOTE: DATA statement used:
      real time           0.22 seconds
 
 
NOTE: The data set WORK.SMOOTH has 31 observations and 5 variables.
NOTE: PROCEDURE LOESS used:
      real time           0.33 seconds
 
 
 
121  %runloess (data=partial,out=smooth,x=weight,y=res2,f=0.5);
 
NOTE: There were 31 observations read from the data set WORK.SMOOTH.
NOTE: PROCEDURE PLOT used:
      real time           0.11 seconds
 
 
NOTE: The data set WORK.SMOOTH has 31 observations and 5 variables.
NOTE: PROCEDURE LOESS used:
      real time           0.10 seconds
 
 
 
122  %runloess (data=partial,out=smooth,x=runtime,y=res3,f=0.5);
 
NOTE: There were 31 observations read from the data set WORK.SMOOTH.
NOTE: PROCEDURE PLOT used:
      real time           0.06 seconds
 
 
NOTE: The data set WORK.SMOOTH has 31 observations and 5 variables.
NOTE: PROCEDURE LOESS used:
      real time           0.26 seconds
 
 
 
123  %runloess (data=partial,out=smooth,x=runpulse,y=res4,f=0.5);
 
NOTE: There were 31 observations read from the data set WORK.SMOOTH.
NOTE: PROCEDURE PLOT used:
      real time           0.00 seconds
 
 
NOTE: The data set WORK.SMOOTH has 31 observations and 5 variables.
NOTE: PROCEDURE LOESS used:
      real time           0.17 seconds
 
 
 
124
 
NOTE: There were 31 observations read from the data set WORK.SMOOTH.
NOTE: PROCEDURE PLOT used:
      real time           0.04 seconds
 
 
125  PROC LOGISTIC DESCENDING DATA = FITNESS;
126       MODEL Y= AGE WEIGHT RUNTIME RUNPULSE NONSMOKE
127    / SELECTION=B SLS=0.1 DETAILS;
128  TITLE2 'BACKWARDS ELIMINATION LOGISTIC REGRESSION WITH ALL VARIABLES';
129
130  RUN;
 
NOTE: PROC LOGISTIC is modeling the probability that Y=1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied in Step 0.
NOTE: Convergence criterion (GCONV=1E-8) satisfied in Step 1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied in Step 2.
NOTE: Convergence criterion (GCONV=1E-8) satisfied in Step 3.
NOTE: There were 31 observations read from the data set WORK.FITNESS.
NOTE: PROCEDURE LOGISTIC used:
      real time           0.60 seconds
 
 
131
132  PROC LOGISTIC DESCENDING DATA=FITNESS;
133    MODEL Y=NONSMOKE RUNTIME / RL rsquare;
134  OUTPUT OUT=OUT1 XBETA=XB STDXBETA=SDXB L=L P=P U=U;
135  TITLE2 'LOGISTIC REGRESSION, REDUCED MODEL';
136
 
NOTE: PROC LOGISTIC is modeling the probability that Y=1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
NOTE: There were 31 observations read from the data set WORK.FITNESS.
NOTE: The data set WORK.OUT1 has 31 observations and 16 variables.
NOTE: PROCEDURE LOGISTIC used:
      real time           0.44 seconds
 
 
137  DATA SETA SETB SETAB; SET FITNESS;
138     UN=UNIFORM(44217);
139       BY=.;  AY=.;
140      IF UN<=0.5 THEN DO; BY=Y; OUTPUT SETB; END;
141      IF UN> 0.5 THEN DO; AY=Y; OUTPUT SETA; END;
142      OUTPUT SETAB;
143
144  *** NOTE REDUCED MODEL WITH NONSMOKE AND RUNTIME ONLY USED FOR VALIDATION
145      DUE TO SMALL SAMPLE SIZES, OTHERWISE DEGENERATE LOGISTIC REGRESSIONS
146      RESULT;
147
 
NOTE: There were 31 observations read from the data set WORK.FITNESS.
NOTE: The data set WORK.SETA has 19 observations and 13 variables.
NOTE: The data set WORK.SETB has 12 observations and 13 variables.
NOTE: The data set WORK.SETAB has 31 observations and 13 variables.
NOTE: DATA statement used:
      real time           0.44 seconds
 
 
148  PROC LOGISTIC DESCENDING DATA=SETAB;
149   MODEL AY=NONSMOKE RUNTIME / RL rsquare;
150    OUTPUT OUT=C L=L P=P U=U;
151  TITLE2 ' LOGISTIC REGRESSION WITH TWO VARIABLES -- A SET';
152
 
NOTE: PROC LOGISTIC is modeling the probability that AY=1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
NOTE: There were 31 observations read from the data set WORK.SETAB.
NOTE: The data set WORK.C has 31 observations and 17 variables.
NOTE: PROCEDURE LOGISTIC used:
      real time           0.26 seconds
 
 
153  data; set;
154  oy=by;
155
156  *****;  %RSQUARE;
 
NOTE: There were 31 observations read from the data set WORK.C.
NOTE: The data set WORK.DATA1 has 31 observations and 18 variables.
NOTE: DATA statement used:
      real time           0.10 seconds
 
 
 
NOTE: There were 31 observations read from the data set WORK.DATA1.
NOTE: The data set WORK.DATA2 has 1 observations and 29 variables.
NOTE: DATA statement used:
      real time           0.05 seconds
 
 
157
158  TITLE2 ' LOGISTIC REGRESSION MODEL VALIDATION -- A SET TO B SET';
159
 
NOTE: There were 1 observations read from the data set WORK.DATA2.
NOTE: PROCEDURE PRINT used:
      real time           0.10 seconds
 
 
160  PROC LOGISTIC DESCENDING DATA=SETAB;
161   MODEL BY=NONSMOKE RUNTIME / RL PPROB=.7 CTABLE;
162    OUTPUT OUT=D L=L P=P U=U;
163  TITLE2 ' LOGISTIC REGRESSION WITH TWO VARIABLES -- B SET';
164
 
NOTE: PROC LOGISTIC is modeling the probability that BY=1.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
NOTE: There were 31 observations read from the data set WORK.SETAB.
NOTE: The data set WORK.D has 31 observations and 17 variables.
NOTE: PROCEDURE LOGISTIC used:
      real time           0.39 seconds
 
 
165  data; set;
166  oy=ay;
167  *****;  %RSQUARE;
 
NOTE: There were 31 observations read from the data set WORK.D.
NOTE: The data set WORK.DATA3 has 31 observations and 18 variables.
NOTE: DATA statement used:
      real time           0.15 seconds
 
 
 
NOTE: There were 31 observations read from the data set WORK.DATA3.
NOTE: The data set WORK.DATA4 has 1 observations and 29 variables.
NOTE: DATA statement used:
      real time           0.00 seconds
 
 
168
169  TITLE2 ' LOGISTIC REGRESSION MODEL VALIDATION -- B SET TO A SET';
170
171  RUN;
 
NOTE: There were 1 observations read from the data set WORK.DATA4.
NOTE: PROCEDURE PRINT used:
      real time           0.11 seconds
 
 
172
173  ods printer close;
NOTE: ODS PRINTER printed 51 pages to c:\sasjobs\stat210\jl2003\sh_blst.pdf.
174  RUN;