You can download the dataset for this chapter by clicking here:
https://stats.idre.ucla.edu/wp-content/uploads/2016/02/dynarski.sas7bdat.
Descriptive statistics and cross-tabulations for key variables. (Not shown in text.)
proc surveymeans data=dynarski;
cluster hhid;
var coll;
weight wt88;
run;
The SURVEYMEANS Procedure
Data Summary
Number of Clusters 3123
Number of Observations 3986
Sum of Weights 1302933368
Statistics
Std Error
Variable Label N Mean of Mean
coll Full-time college by age 23 3986 0.494350 0.010515
Statistics
Variable 95% CL for Mean
coll 0.47373264 0.51496809
proc freq data=dynarski;
table fatherdec*yearsr;
run;
The FREQ Procedure
Table of fatherdec by yearsr
fatherdec(Father deceased by age 18)
yearsr(Year in which a senior)
Frequency
Percent
Row Pct
Col Pct 79 80 81 82 83 Total
Father not decea 892 986 867 828 222 3795
sed 22.38 24.74 21.75 20.77 5.57 95.21
23.50 25.98 22.85 21.82 5.85
95.61 95.73 94.34 95.28 94.47
Father deceased 41 44 52 41 13 191
1.03 1.10 1.30 1.03 0.33 4.79
21.47 23.04 27.23 21.47 6.81
4.39 4.27 5.66 4.72 5.53
Total 933 1030 919 869 235 3986
23.41 25.84 23.06 21.80 5.90 100.00
(a) Direct Estimate shown in Table 8.1 on page
143. The means of coll are shown in the rows labeled _subpop_3 and
_subpop_4. This output also replicates part of the table of variable means and differences from the Dynarski article referenced in the chapter
(not shown in the text).
proc sort data=dynarski;
by fatherdec offer;
run;
proc surveymeans data=dynarski;
domain fatherdec*offer;
cluster hhid;
var coll;
weight wt88;
run;
The SURVEYMEANS Procedure
Data Summary
Number of Clusters 3123
Number of Observations 3986
Sum of Weights 1302933368
Statistics
Std Error
Variable Label N Mean of Mean
coll Full-time college by age 23 3986 0.494350 0.010515
Statistics
Variable 95% CL for Mean
coll 0.47373264 0.51496809
Domain Analysis: Father deceased by age 18*Senior in year SSSB support available
Senior in
year SSSB
Father deceased support
by age 18 available Variable Label N Mean
Father not deceased 0 coll Full-time college by age 23 1050 0.475694
1 coll Full-time college by age 23 2745 0.501702
Father deceased 0 coll Full-time college by age 23 54 0.352218
1 coll Full-time college by age 23 137 0.560456
Domain Analysis: Father deceased by age 18*Senior in year SSSB support available
Senior in
year SSSB
Father deceased support Std Error
by age 18 available Variable of Mean 95% CL for Mean
Father not deceased 0 coll 0.018865 0.43870461 0.51268248
1 coll 0.012174 0.47783269 0.52557057
Father deceased 0 coll 0.081245 0.19291968 0.51151598
1 coll 0.052744 0.45703939 0.66387184
(b) Linear-Probability Model (OLS) Estimate shown in Table 8.1 on page 143.
proc surveyreg data=dynarski;
model coll = offer;
cluster hhid;
weight wt88;
where fatherdec=1;
run;
The SURVEYREG Procedure
Regression Analysis for Dependent Variable coll
Data Summary
Number of Observations 191
Sum of Weights 51656801
Weighted Mean of coll 0.49986
Weighted Sum of coll 25821369
Design Summary
Number of Clusters 172
Fit Statistics
R-square 0.03578
Root MSE 0.4936
Denominator DF 171
Tests of Model Effects
Effect Num DF F Value Pr > F
Model 1 4.93 0.0277
Intercept 1 18.59 <.0001
offer 1 4.93 0.0277
NOTE: The denominator degrees of freedom for the F tests is 171.
Estimated Regression Coefficients
Standard
Parameter Estimate Error t Value Pr > |t|
Intercept 0.35221783 0.08168396 4.31 <.0001
offer 0.20823778 0.09375011 2.22 0.0277
Figure 8.1 on page 155. Note that some of the output from the following section of code is not shown.
data dynarski;
set dynarski;
if offer=1 then post = 0;
if offer=0 then post = 1;
run;
proc sort data=dynarski;
by fatherdec post;
run;
proc surveymeans data=dynarski;
by fatherdec post;
cluster hhid;
var coll;
weight wt88;
ods output statistics = table81;
run;
proc print data=table81;
run;
proc format;
value post 0="pre-1981"
1="post-1981";
run;
proc sgplot data=table81;
by fatherdec;
series x=post y=mean ;
yaxis min=.3 max=.6;
xaxis tickvalueformat=post. values=(0 1);
run;
Table 8.2 on page 157, labeled "(First Diff)". (Note this
replicates the last regression model.)
proc surveyreg data=dynarski;
model coll = offer;
cluster hhid;
weight wt88;
where fatherdec = 1 ;
run;
The SURVEYREG Procedure
Regression Analysis for Dependent Variable coll
Data Summary
Number of Observations 191
Sum of Weights 51656801
Weighted Mean of coll 0.49986
Weighted Sum of coll 25821369
Design Summary
Number of Clusters 172
Fit Statistics
R-square 0.03578
Root MSE 0.4936
Denominator DF 171
Tests of Model Effects
Effect Num DF F Value Pr > F
Model 1 4.93 0.0277
Intercept 1 18.59 <.0001
offer 1 4.93 0.0277
NOTE: The denominator degrees of freedom for the F tests is 171.
Estimated Regression Coefficients
Standard
Parameter Estimate Error t Value Pr > |t|
Intercept 0.35221783 0.08168396 4.31 <.0001
offer 0.20823778 0.09375011 2.22 0.0277
Table 8.2 on page 157, labeled "(Second Diff)".
proc surveyreg data=dynarski;
model coll = offer;
cluster hhid;
weight wt88;
where fatherdec = 0;
run;
The SURVEYREG Procedure
Regression Analysis for Dependent Variable coll
Data Summary
Number of Observations 3795
Sum of Weights 1251276567
Weighted Mean of coll 0.49412
Weighted Sum of coll 618284217
Design Summary
Number of Clusters 2984
Fit Statistics
R-square 0.000559
Root MSE 0.5000
Denominator DF 2983
Tests of Model Effects
Effect Num DF F Value Pr > F
Model 1 1.50 0.2215
Intercept 1 635.66 <.0001
offer 1 1.50 0.2215
NOTE: The denominator degrees of freedom for the F tests is 2983.
Estimated Regression Coefficients
Standard
Parameter Estimate Error t Value Pr > |t|
Intercept 0.47569354 0.01886756 25.21 <.0001
offer 0.02600809 0.02126729 1.22 0.2215
Table 8.4 on page 161.
proc surveyreg data=dynarski;
model coll = offer fatherdec offer*fatherdec;
cluster hhid;
weight wt88;
estimate offer 1 offer*fatherdec 1 ;
run;
The SURVEYREG Procedure
Regression Analysis for Dependent Variable coll
Data Summary
Number of Observations 3986
Sum of Weights 1302933368
Weighted Mean of coll 0.49435
Weighted Sum of coll 644105586
Design Summary
Number of Clusters 3123
Fit Statistics
R-square 0.001961
Root MSE 0.4997
Denominator DF 3122
Tests of Model Effects
Effect Num DF F Value Pr > F
Model 3 2.19 0.0875
Intercept 1 635.36 <.0001
offer 1 1.49 0.2216
fatherdec 1 2.19 0.1391
offer*fatherdec 1 3.61 0.0574
NOTE: The denominator degrees of freedom for the F tests is 3122.
Estimated Regression Coefficients
Standard
Parameter Estimate Error t Value Pr > |t|
Intercept 0.4756935 0.01887204 25.21 <.0001
offer 0.0260081 0.02127234 1.22 0.2216
fatherdec -0.1234757 0.08345651 -1.48 0.1391
offer*fatherdec 0.1822297 0.09587712 1.90 0.0574
The SURVEYREG Procedure
Regression Analysis for Dependent Variable coll
Estimate
Standard
Label Estimate Error DF t Value Pr > |t|
Row 1 0.2082 0.09328 3122 2.23 0.0257