UCLA Academic Technology Services HomeServicesClassesContactJobs
Search

SUDAAN FAQ 
How can I do regression estimation with survey data?

This example is taken from Lehtonen and Pahkinen's Practical Methods for Design and Analysis of Complex Surveys.
page 107 Table 3.14  Model-assisted estimation results for the population total of ue91 from an SRS sample of eight elements drawn from the Province'91 population.
data page106;
  input id str clu wt ue91 meanz hou85 diffhou85 smplrat;
  fpc = 32;
  cards;
  1 1 1 4 4123 2867 26881 -24014 .25
  2 1 4 4 760 2867 4896 -2029 .25
  3 1 5 4 721 2867 3730 -863 .25
  4 1 15 4 142 2867 556 2311 .25
  5 1 18 4 187 2867 1463 1404 .25
  6 1 26 4 331 2867 1946 921 .25
  7 1 30 4 127 2867 834 2033 .25
  8 1 31 4 219 2867 932 1935 .25
  ;
run;
strategy:  design-based estimator with srs
proc descript data = page106 filetype = sas design = wor totals deft;
  weight wt;
  nest _one_;
  totcnt fpc;
  var ue91;
run;
Number of observations read    :      8    Weighted count :       32
Denominator degrees of freedom :      7
Variance Estimation Method: Taylor Series (WOR)
by: Variable, One.

-----------------------------------------------------
|                 |                  |
| Variable        |                  | One
|                 |                  | 1            |
-----------------------------------------------------
|                 |                  |              |
| UE91            | Sample Size      |            8 |
|                 | Weighted Size    |        32.00 |
|                 | Total            |     26440.00 |
|                 | SE Total         |     13282.26 |
|                 | Mean             |       826.25 |
|                 | SE Mean          |       415.07 |
|                 | DEFF Mean #4     |         0.75 |
|                 | DEFF Total #4    |         0.75 |
-----------------------------------------------------
strategy:  poststratified estimator with srs*pos
proc descript data = page106 filetype = sas design = wor totals;
  weight wt;
  nest _one_;
  totcnt fpc;
  var ue91 hou85;
  subgroup str;
  levels 1;
  postvar str;
  postwgt 8;
run;
Number of observations read    :      8    Weighted count :       32
Denominator degrees of freedom :      7
Variance Estimation Method: Taylor Series (WOR)
Post-stratified estimates
by: Variable, STR.

--------------------------------------------------------------------
|                 |                  |
| Variable        |                  | STR
|                 |                  | Total        | 1            |
--------------------------------------------------------------------
|                 |                  |              |              |
| UE91            | Sample Size      |            8 |            8 |
|                 | Weighted Size    |         8.00 |         8.00 |
|                 | Total            |      6610.00 |      6610.00 |
|                 | SE Total         |      3320.56 |      3320.56 |
|                 | Mean             |       826.25 |       826.25 |
|                 | SE Mean          |       415.07 |       415.07 |
--------------------------------------------------------------------
|                 |                  |              |              |
| HOU85           | Sample Size      |            8 |            8 |
|                 | Weighted Size    |         8.00 |         8.00 |
|                 | Total            |     41238.00 |     41238.00 |
|                 | SE Total         |     21824.64 |     21824.64 |
|                 | Mean             |      5154.75 |      5154.75 |
|                 | SE Mean          |      2728.08 |      2728.08 |
--------------------------------------------------------------------
strategy:  ratio estimator with srs*rat

This code is shown above for page 102.
strategy:  regression estimator with srs*reg

The code below produces the estimate of b-hat, 0.152, shown in the middle of page 106.
proc regress data = page106 filetype = sas design = wor;
  weight wt;
  nest _one_;
  totcnt fpc;
  model ue91 = hou85;
  setenv decwidth = 3;
run;
Number of observations read       :      8    Weighted count:       32
Observations used in the analysis :      8    Weighted count:       32
Denominator degrees of freedom    :      7

Maximum number of estimable parameters for the model is  2

File PAGE106 contains    8 Clusters
   8 clusters were used to fit the model
Maximum cluster size is   1 records
Minimum cluster size is   1 records

Weighted mean response is 826.250000

Multiple R-Square for the dependent variable UE91: 0.998249
Variance Estimation Method: Taylor Series (WOR)
SE Method: Robust (Binder, 1983)
Working Correlations: Independent
Link Function: Identity
Response variable UE91: UE91

----------------------------------------------------------------------
Independent                                                   P-value
  Variables and        Beta                                   T-Test
  Effects              Coeff.          SE Beta   T-Test B=0   B=0
----------------------------------------------------------------------
Intercept                  42.655       20.540        2.077      0.076
HOU85                       0.152        0.001      212.012      0.000
----------------------------------------------------------------------
-------------------------------------------------------

Contrast               Degrees
                       of                      P-value
                       Freedom        Wald F   Wald F
-------------------------------------------------------
OVERALL MODEL             2.000   102820.497      0.000
MODEL MINUS
  INTERCEPT               1.000    44949.184      0.000
INTERCEPT                 1.000        4.312      0.076
HOU85                     1.000    44949.184      0.000
-------------------------------------------------------
This gives the estimate of the total of hou85, 164952, which is needed for the equation.  Note that 32*2867 = 91753.
proc descript data = page106 filetype = sas design = wor totals;
  weight wt;
  nest _one_;
  totcnt fpc;
  var hou85;
run;
Number of observations read    :      8    Weighted count :       32
Denominator degrees of freedom :      7
Variance Estimation Method: Taylor Series (WOR)
by: Variable, One.

-----------------------------------------------------
|                 |                  |
| Variable        |                  | One
|                 |                  | 1            |
-----------------------------------------------------
|                 |                  |              |
| HOU85           | Sample Size      |            8 |
|                 | Weighted Size    |        32.00 |
|                 | Total            |    164952.00 |
|                 | SE Total         |     87298.57 |
|                 | Mean             |      5154.75 |
|                 | SE Mean          |      2728.08 |
-----------------------------------------------------

How to cite this page

Report an error on this page

UCLA Researchers are invited to our Statistical Consulting Services
We recommend others to our list of Other Resources for Statistical Computing Help
These pages are Copyrighted (c) by UCLA Academic Technology Services


The content of this web site should not be construed as an endorsement of any particular web site, book, or software product by the University of California.