1          OPTIONS NONOTES NOSTIMER NOSOURCE NOSYNTAXCHECK;
 72         
 73         /* multilogit.sas */
 74         title2 'Multinomial Logit Model';
 75         %include '/folders/myfolders/441s18/Lecture/mathread2.sas';
 NOTE: Format YNFMT is already on the library WORK.FORMATS.
 NOTE: Format YNFMT has been output.
 NOTE: Format CRSFMT is already on the library WORK.FORMATS.
 NOTE: Format CRSFMT has been output.
 NOTE: Format NFMT is already on the library WORK.FORMATS.
 NOTE: Format NFMT has been output.
 
 NOTE: PROCEDURE FORMAT used (Total process time):
       real time           0.00 seconds
       cpu time            0.00 seconds
       
 
 196        /* mathread2.sas creates missused (any of hsgpa hscalc precalc missing)
 197           and outcome (Pass-Fail-Gone)*/
 198        
 
 NOTE: The infile '/folders/myfolders/441s18/Lecture/exploremath.data.txt' is:
       Filename=/folders/myfolders/441s18/Lecture/exploremath.data.txt,
       Owner Name=root,Group Name=vboxsf,
       Access Permission=-rwxrwx---,
       Last Modified=January 18, 2016 17:34:49,
       File Size (bytes)=44583
 
 NOTE: 579 records were read from the infile '/folders/myfolders/441s18/Lecture/exploremath.data.txt'.
       The minimum record length was 75.
       The maximum record length was 75.
 NOTE: Missing values were generated as a result of performing an operation on missing values.
       Each place is given by: (Number of times) at (Line):(Column).
       99 at 112:24    99 at 113:18    142 at 183:14   204 at 183:21   
 NOTE: The data set WORK.MATHEX has 579 observations and 36 variables.
 NOTE: DATA statement used (Total process time):
       real time           0.02 seconds
       cpu time            0.01 seconds
       
 199        proc freq data=mathex;
 
 200             title3 'Outcome by passed just to check';
 201             tables outcome*passed
 202                    / norow nocol nopercent missing;
 203        run;
 
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE FREQ used (Total process time):
       real time           0.10 seconds
       cpu time            0.09 seconds
       
 
 204        
 205        proc freq data=mathex;
 206             title3 'Missused by Outcome with proc freq';
 207             tables missused * outcome / nocol nopercent chisq;
 208        run;
 
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE FREQ used (Total process time):
       real time           0.08 seconds
       cpu time            0.07 seconds
       
 
 209        
 210        /* Multinomial Logit model for missused by outcome is
 211        
 212             ln(pi1/pi3) = beta01 + beta11 x             Fail vs. Pass
 213             ln(pi2/pi3) = beta02 + beta12 x             Gone vs. Pass
 214        */
 215        
 216        
 217        proc logistic data=mathex outest=ParmNames;
 218             title3 'Misssused by Outcome with proc logistic';
 219             model outcome (ref='Pass') = missused / link = glogit;
 220             contrast 'Hsgpa, hscalc or precalc missing' missused 1;
 221        run;
 
 NOTE: PROC LOGISTIC is fitting the generalized logit model. The logits modeled contrast each response category against the 
       reference category (outcome='Pass').
 NOTE: Convergence criterion (GCONV=1E-8) satisfied.
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: The data set WORK.PARMNAMES has 1 observations and 10 variables.
 NOTE: PROCEDURE LOGISTIC used (Total process time):
       real time           0.18 seconds
       cpu time            0.16 seconds
       
 
 222        
 223        /* Find out the parameter names, written with outest. */
 224        
 225        proc transpose data=ParmNames;
 226        run;
 
 NOTE: There were 1 observations read from the data set WORK.PARMNAMES.
 NOTE: The data set WORK.DATA2 has 5 observations and 3 variables.
 NOTE: PROCEDURE TRANSPOSE used (Total process time):
       real time           0.01 seconds
       cpu time            0.01 seconds
       
 
 227        proc print;
 228        run;
 
 NOTE: There were 5 observations read from the data set WORK.DATA2.
 NOTE: PROCEDURE PRINT used (Total process time):
       real time           0.04 seconds
       cpu time            0.04 seconds
       
 
 229        
 230        proc logistic data = mathex;
 231             title3 'Contrast versus test';
 232             model outcome (ref='Pass') = missused / link = glogit;
 233             contrast 'Missused method 1' missused 1;
 234             MissusedMethod2: test missused_Fail = missused_Gone = 0;
 235        run;
 
 NOTE: PROC LOGISTIC is fitting the generalized logit model. The logits modeled contrast each response category against the 
       reference category (outcome='Pass').
 NOTE: Convergence criterion (GCONV=1E-8) satisfied.
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE LOGISTIC used (Total process time):
       real time           0.19 seconds
       cpu time            0.18 seconds
       
 
 236        
 237        proc iml;
 NOTE: IML Ready
 238             title3 'Estimate Probabilities using output from proc logistic';
 239              b01 = -1.4123;
 239      !                      b11 = 0.6461;
 239      !                                     /* Fail */
 240              b02 = -1.0245;
 240      !                      b12 = 1.3670;
 240      !                                     /* Gone */
 241              missused = 0;
 242              L1 = b01 + b11*missused;
 243              L2 = b02 + b12*missused;
 244              denom = 1 + exp(L1) + exp(L2);
 245              Fail = exp(L1)/denom;
 245      !                             Gone = exp(L2)/denom;
 245      !                                                   Pass = 1/denom;
 246              print "Not missing:" Fail Gone Pass;
 247              missused = 1;
 248              L1 = b01 + b11*missused;
 249              L2 = b02 + b12*missused;
 250              denom = 1 + exp(L1) + exp(L2);
 251              Fail = exp(L1)/denom;
 251      !                             Gone = exp(L2)/denom;
 251      !                                                   Pass = 1/denom;
 252              print "Yes Missing:" Fail Gone Pass;
 253        
 NOTE: Exiting IML.
 NOTE: PROCEDURE IML used (Total process time):
       real time           0.08 seconds
       cpu time            0.07 seconds
       
 254        proc freq  data = mathex;
 
 255             title3 'Missused by outcome again for comparison';
 256             tables missused * outcome / nocol nopercent;
 257        
 258        /* Lots of exploration is not shown here. */
 259        
 
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE FREQ used (Total process time):
       real time           0.09 seconds
       cpu time            0.08 seconds
       
 260        proc logistic data = mathex;
 
 261             title3 'hsgpa hscalc precalc mtongue';
 262             model outcome (ref='Pass') =
 263                           hsgpa hscalc precalc mtongue / link = glogit;
 264        run;
 
 NOTE: PROC LOGISTIC is fitting the generalized logit model. The logits modeled contrast each response category against the 
       reference category (outcome='Pass').
 NOTE: Convergence criterion (GCONV=1E-8) satisfied.
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE LOGISTIC used (Total process time):
       real time           0.20 seconds
       cpu time            0.19 seconds
       
 
 265        
 266        /* Allowing for academic background, students whose first language is English
 267        are more likely to fail the course as opposed to passing, and less likely to
 268        disappear as opposed to passing. If this is replicated, it will be very
 269        interesting. Explore further.
 270        
 271           We want to know whether failing is different from disappearing in terms
 272           of their relationship to the explanatory variables. We are getting
 273           advanced here. What is H0?
 274        
 275           Recall the response categories are 1=Fail 2=Gone 3=Pass.
 276        
 277        Model (using b instead of beta) is
 278        
 279             ln(pi1/pi3) = b01 + b11 hsgpa + b21 hscalc + b31 precalc + b41 mtongue
 280             ln(pi2/pi3) = b02 + b12 hsgpa + b22 hscalc + b32 precalc + b42 mtongue
 281        
 282             The null hypothesis is b11=b12, b21=b22, b31=b32, b41=b42
 283        
 284        Parameter names are easy to guess.  */
 285        
 286        proc logistic data = mathex;
 287             title3 'Different coefficients for Gone and Fail?';
 288             model outcome (ref='Pass') = hsgpa hscalc precalc mtongue / link = glogit;
 289             DiffOverall: test hsgpa_Fail = hsgpa_Gone, hscalc_Fail = hscalc_Gone,
 290                               precalc_Fail = precalc_Gone, mtongue_Fail = mtongue_Gone;
 291             Diff_hsgpa:   test hsgpa_Fail = hsgpa_Gone;
 292             Diff_hscalc:  test hscalc_Fail = hscalc_Gone;
 293             Diff_precalc: test precalc_Fail = precalc_Gone;
 294             Diff_mtongue: test mtongue_Fail = mtongue_Gone;
 295        run;
 
 NOTE: PROC LOGISTIC is fitting the generalized logit model. The logits modeled contrast each response category against the 
       reference category (outcome='Pass').
 NOTE: Convergence criterion (GCONV=1E-8) satisfied.
 NOTE: There were 579 observations read from the data set WORK.MATHEX.
 NOTE: PROCEDURE LOGISTIC used (Total process time):
       real time           0.24 seconds
       cpu time            0.24 seconds
       
 
 296        
 297        /************************** Replication ***********************
 298        For interpretation, want to replicate 8 findings:
 299        Gone vs. Pass and Fail vs. Pass for each explanatory variable.
 300        ***************************************************************/
 301        
 302        %include '/folders/myfolders/441s18/Lecture/readreplic.sas';
 NOTE: Format YNFMT is already on the library WORK.FORMATS.
 NOTE: Format YNFMT has been output.
 NOTE: Format CRSFMT is already on the library WORK.FORMATS.
 NOTE: Format CRSFMT has been output.
 NOTE: Format NFMT is already on the library WORK.FORMATS.
 NOTE: Format NFMT has been output.
 
 NOTE: PROCEDURE FORMAT used (Total process time):
       real time           0.00 seconds
       cpu time            0.00 seconds
       
 
 416        if (0<=mark<=49) then outcome = 'Fail';
 417           else if (50<=mark<=100) then outcome = 'Pass';
 418           else outcome = 'Gone';
 419        
 
 NOTE: The infile '/folders/myfolders/441s18/Lecture/replicmath.data.txt' is:
       Filename=/folders/myfolders/441s18/Lecture/replicmath.data.txt,
       Owner Name=root,Group Name=vboxsf,
       Access Permission=-rwxrwx---,
       Last Modified=February 04, 2018 14:39:27,
       File Size (bytes)=38214
 
 NOTE: 579 records were read from the infile '/folders/myfolders/441s18/Lecture/replicmath.data.txt'.
       The minimum record length was 64.
       The maximum record length was 64.
 NOTE: Missing values were generated as a result of performing an operation on missing values.
       Each place is given by: (Number of times) at (Line):(Column).
       81 at 339:24   81 at 340:18   
 NOTE: The data set WORK.MATHREP has 579 observations and 35 variables.
 NOTE: DATA statement used (Total process time):
       real time           0.02 seconds
       cpu time            0.01 seconds
       
 420        proc logistic data = mathrep;
 
 421             title2 'Replicate hsgpa hscalc precalc calc mtongue 0.05/8 = .00625';
 422             model outcome (ref='Pass') = hsgpa hscalc precalc mtongue / link = glogit;
 423             Diff_mtongue: test mtongue_Fail = mtongue_Gone;
 424        
 425        
 426        
 427        OPTIONS NONOTES NOSTIMER NOSOURCE NOSYNTAXCHECK;
 440