/* Data for 2007-2008 Economics Graduates */

Data subset1;
input smugpa subsetgpa;
datalines;
3.455	3.925
3.577	3.925
3.540	3.825
3.713	3.675
3.640	3.650
3.026	3.500
3.060	3.500
3.225	3.425
2.912	3.425
3.228	3.275
3.102	3.250
3.061	3.250
3.277	3.250
2.872	3.100
3.486	3.075
2.573	3.075
2.368	3.075
3.340	3.075
2.735	3.075
2.943	3.075
3.000	3.075
3.397	3.025
2.195	3.000
2.947	3.000
3.142	3.000
3.436	2.925
2.748	2.925
2.826	2.925
3.095	2.850
3.045	2.825
3.251	2.825
2.983	2.825
2.933	2.825
3.377	2.800
2.809	2.750
3.216	2.750
2.297	2.675
2.729	2.675
2.893	2.675
2.488	2.675
3.201	2.650
2.377	2.575
2.651	2.540
2.976	2.525
2.785	2.500
2.732	2.500
2.927	2.475
2.453	2.425
2.897	2.425
2.646	2.425
2.540	2.400
2.642	2.400
1.650	2.400
3.243	2.400
2.390	2.325
2.531	2.250
2.321	2.250
2.276	2.200
2.619	2.200
2.741	2.175
2.610	2.175
3.250	2.175
2.506	2.150
2.326	2.100
2.228	2.100
2.760	2.000
2.443	1.540
2.421	1.280
2.390	1.200
;

Proc reg data = subset1;
  model smugpa = subsetgpa;
  output r=resid out=results1;
  title 'Regression for 2007-2008 Grads';
  run;

/* Data for 2008-2009 Economics Graduates */
Data subset2;
   input smugpa subsetgpa;
datalines;
3.941	4.000
3.669	3.925
3.487	3.925
3.394	3.825
3.260	3.825
3.590	3.750
3.317	3.750
3.502	3.675
3.644	3.675
3.505	3.575
2.930	3.350
3.277	3.350
2.612	3.325
3.351	3.325
2.694	3.250
3.130	3.150
3.039	3.100
2.683	3.100
2.918	3.075
3.306	3.000
3.085	3.000
2.928	2.900
3.417	2.900
2.986	2.825
2.916	2.750
3.208	2.750
2.449	2.750
2.865	2.750
3.079	2.750
2.314	2.680
3.094	2.675
3.002	2.675
2.652	2.675
2.551	2.600
3.005	2.600
2.945	2.575
2.730	2.575
2.725	2.550
2.229	2.525
3.318	2.500
2.169	2.425
2.748	2.425
2.865	2.425
2.530	2.425
2.490	2.400
3.166	2.350
2.729	2.260
2.512	2.250
2.549	2.200
2.153	1.820
;

Proc reg data = subset2;
  model smugpa = subsetgpa;
  output r=resid out=results2;
  title 'Regression for 2008-2009 Grads';
  run;

  quit;

  /* Do White's test for Heteroskedasticity for each of
     the group regression equations */

  Data results1;
    set results1;
    res2 = resid**2;
	subsetgpa2 = subsetgpa**2;

  Data results2;
    set results2;
	res2 = resid**2;
	subsetgpa2 = subsetgpa**2;

 Proc reg data = results1;
    model res2 = subsetgpa subsetgpa2;
	title 'Whites test for Heteroskedasticity in Group 1'; 
	run;

Proc reg data = results2;
    model res2 = subsetgpa subsetgpa2;
	title 'Whites test for Heteroskedasticity in Group 2'; 
	run;

/*  We can conclude that neither of the group equations
	is hindered by heteroskedasticity.  Therefore,
	the OLS estimates are valid estimates for the
	regression parameters of the linear regressions
	involving the dependent variable SMUGPA and the
	explanatory variable SUBSETGPA.  In both cases the
	SUBSETGPA variable is a statistically significant
	explanator of the variation in SMU cumulative GPAs
	among graduating senior economics majors.  */