use https://stats.idre.ucla.edu/stat/stata/examples/methods_matter/chapter8/dynarski, clear
svyset, clear
svyset [pw=wt88], psu(hhid)
pweight: wt88
VCE: linearized
Single unit: missing
Strata 1: <one>
SU 1: hhid
FPC 1: <zero>
Descriptive statistics and cross-tabulations for key variables. (Not shown in text.)
svy: mean coll
(running mean on estimation sample)
Survey: Mean estimation
Number of strata = 1 Number of obs = 3986
Number of PSUs = 3123 Population size = 1302933368
Design df = 3122
--------------------------------------------------------------
| Linearized
| Mean Std. Err. [95% Conf. Interval]
-------------+------------------------------------------------
coll | .4943504 .0105154 .4737326 .5149681
--------------------------------------------------------------
tabulate fatherdec yearsr
Father deceased by | Year in which a senior
age 18 | 79 80 81 82 83 | Total
--------------------+-------------------------------------------------------+----------
Father not deceased | 892 986 867 828 222 | 3,795
Father deceased | 41 44 52 41 13 | 191
--------------------+-------------------------------------------------------+----------
Total | 933 1,030 919 869 235 | 3,986
(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).
svy: mean coll, over(fatherdec offer)
(running mean on estimation sample)
Survey: Mean estimation
Number of strata = 1 Number of obs = 3986
Number of PSUs = 3123 Population size = 1302933368
Design df = 3122
Over: fatherdec offer
_subpop_1: Father not deceased 0
_subpop_2: Father not deceased 1
_subpop_3: Father deceased 0
_subpop_4: Father deceased 1
--------------------------------------------------------------
| Linearized
Over | Mean Std. Err. [95% Conf. Interval]
-------------+------------------------------------------------
coll |
_subpop_1 | .4756935 .0188649 .4387046 .5126825
_subpop_2 | .5017016 .0121735 .4778327 .5255706
_subpop_3 | .3522178 .0812446 .1929197 .511516
_subpop_4 | .5604556 .0527439 .4570394 .6638718
-------------+------------------------------------------------
(b) Linear-Probability Model (OLS) Estimate shown in Table 8.1 on page 143.
svy if fatherdec==1 : regress coll offer
(running regress on estimation sample)
Survey: Linear regression
Number of strata = 1 Number of obs = 191
Number of PSUs = 172 Population size = 51656801
Design df = 171
F( 1, 171) = 4.96
Prob > F = 0.0272
R-squared = 0.0358
------------------------------------------------------------------------------
| Linearized
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
offer | .2082378 .0935031 2.23 0.027 .0236689 .3928067
_cons | .3522178 .0814687 4.32 0.000 .191404 .5130317
------------------------------------------------------------------------------
Figure 8.1 on page 155.
recode offer (0=1) (1=0), gen(post)
twoway lfit coll post [pw=wt88], by(fatherdec, note("")) ///
ytitle("P{college}") yscale(range(.3 .6)) ///
xtitle("Year") xlabel(0 "Pre-1981" 1 "Post-1981") xscale(range(-.25 1.25))
Table 8.2 on page 157, labeled “(First Diff)”. (Note this
replicates the last regression model.)
svy if fatherdec==1 : regress coll offer
(running regress on estimation sample)
Survey: Linear regression
Number of strata = 1 Number of obs = 191
Number of PSUs = 172 Population size = 51656801
Design df = 171
F( 1, 171) = 4.96
Prob > F = 0.0272
R-squared = 0.0358
------------------------------------------------------------------------------
| Linearized
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
offer | .2082378 .0935031 2.23 0.027 .0236689 .3928067
_cons | .3522178 .0814687 4.32 0.000 .191404 .5130317
------------------------------------------------------------------------------
Table 8.2 on page 157, labeled “(Second Diff)”.
svy if fatherdec==0 : regress coll offer
(running regress on estimation sample)
Survey: Linear regression
Number of strata = 1 Number of obs = 3795
Number of PSUs = 2984 Population size = 1251276567
Design df = 2983
F( 1, 2983) = 1.50
Prob > F = 0.2214
R-squared = 0.0006
------------------------------------------------------------------------------
| Linearized
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
offer | .0260081 .0212645 1.22 0.221 -.0156865 .0677026
_cons | .4756935 .0188651 25.22 0.000 .4387037 .5126834
------------------------------------------------------------------------------
Table 8.4 on page 161.
svy: regress coll offer fatherdec offerxfatherdec
(running regress on estimation sample)
Survey: Linear regression
Number of strata = 1 Number of obs = 3986
Number of PSUs = 3123 Population size = 1302933368
Design df = 3122
F( 3, 3120) = 2.19
Prob > F = 0.0875
R-squared = 0.0020
------------------------------------------------------------------------------
| Linearized
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
offer | .0260081 .0212643 1.22 0.221 -.0156854 .0677016
fatherdec | -.1234757 .0834251 -1.48 0.139 -.2870493 .0400979
offerxfath~c | .1822297 .095841 1.90 0.057 -.0056881 .3701475
_cons | .4756935 .0188649 25.22 0.000 .4387046 .5126825
------------------------------------------------------------------------------
The “first difference” for cases where the father is not deceased shown in
Table 8.2 can be seen here as the coefficient for offer. The term for the
interaction (offerxfatherdec) is the “second difference.” The syntax
below calculates the “first difference” for cases where the father is deceased.
lincom offer + offerxfatherdec
( 1) offer + offerxfatherdec = 0
------------------------------------------------------------------------------
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | .2082378 .0932458 2.23 0.026 .0254085 .3910671
------------------------------------------------------------------------------
Table 8.4 on page 161 can
also be reproduced using the factor variable syntax introduced in Stata 11.
svy: regress coll offer##fatherdec
(running regress on estimation sample)
Survey: Linear regression
Number of strata = 1 Number of obs = 3986
Number of PSUs = 3123 Population size = 1302933368
Design df = 3122
F( 3, 3120) = 2.19
Prob > F = 0.0875
R-squared = 0.0020
------------------------------------------------------------------------------
| Linearized
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
1.offer | .0260081 .0212643 1.22 0.221 -.0156854 .0677016
1.fatherdec | -.1234757 .0834251 -1.48 0.139 -.2870493 .0400979
|
offer#|
fatherdec |
1 1 | .1822297 .095841 1.90 0.057 -.0056881 .3701475
|
_cons | .4756935 .0188649 25.22 0.000 .4387046 .5126825
------------------------------------------------------------------------------
lincom 1.offer + 1.offer#1.father
( 1) 1.offer + 1.offer#1.fatherdec = 0
------------------------------------------------------------------------------
coll | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | .2082378 .0932458 2.23 0.026 .0254085 .3910671
------------------------------------------------------------------------------