From GAGNIERE Bertrand 172658 To "'statalist@hsphsun2.harvard.edu'" Subject st: Adjusted means Date Mon, 21 Oct 2002 15:50:35 +0200

```Dear all,

I am working on a descriptive survey about nutritional habits in Ukraine.
This survey was carried out in two different seasons and in different places
where levels of contamination following the Chernobyl accident varied.
My first goal is to provide detailled tables in order to describe the
intakes by age, level of contamination and season. The following example
concerns salofat

. table age period zone, c(mean salofat sd salofat) format(%9.3f) center

----------------------------------------------------------------------------
----------------------------------------------
|                                      level of contamination
and period
| --------- <1 Ci --------    -------- 1-5 Ci --------
-------- 5-15 Ci -------    -------- >15 Ci --------
Age | spring 2000   fall 2001     spring 2000   fall 2001     spring
2000   fall 2001     spring 2000   fall 2001
------------+---------------------------------------------------------------
----------------------------------------------
18-29 years |    47.000                      51.818       35.000
53.125       87.500         45.357       52.143
|    21.095        0.000         25.620       17.321
22.510       38.891         18.341       21.185
|
30-39 years |    38.125                      57.143       56.667
71.818       59.286         35.769       24.167
|    26.449        0.000         29.277        5.774
37.899       19.242         18.802       21.866
|
40-59 years |    55.000                      58.000       75.000
60.769       55.417         53.971       37.143
|    25.071        0.000         22.345       37.796
28.929       11.958         28.326       23.077
|
60-74 years |    50.833                      46.364       58.333
53.654       70.000         95.000       51.000
|    18.005        0.000         28.468       11.690
22.607       39.306         63.640       24.083
|
75 + |    50.000                      40.000
62.500                      50.000
|     7.071        0.000         12.649        0.000
34.776                                    0.000
----------------------------------------------------------------------------
----------------------------------------------

Note that in this example, nobody has eaten salofat in fall 2001 in the less
contaminated region (< 1Ci).

In a second step, I need to provide adjusted means. Here is what I did:

xi: regress salofat period i.age i.zone
i.age             _Iage_1-5           (_Iage_1 for age==18-29 years omitted)
i.zone            _Izone_0-4          (naturally coded; _Izone_0 omitted)

Source |       SS       df       MS              Number of obs =
271
-------------+------------------------------           F(  8,   262) =
2.86
Model |  15878.8172     8  1984.85216           Prob > F      =
0.0046
Residual |  182138.019   262  695.183277           R-squared     =
0.0802
0.0521
Total |  198016.836   270  733.395688           Root MSE      =
26.366

----------------------------------------------------------------------------
--
salofat |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
period |   -.958352   3.642396    -0.26   0.793    -8.130448
6.213744
_Iage_2 |  -3.164714   5.099201    -0.62   0.535    -13.20535
6.875917
_Iage_3 |   4.166338   4.687264     0.89   0.375    -5.063164
13.39584
_Iage_4 |   2.707066   5.308515     0.51   0.611    -7.745717
13.15985
_Iage_5 |  -2.127529   7.024452    -0.30   0.762    -15.95909
11.70404
_Izone_2 |   5.593309   5.739183     0.97   0.331    -5.707485
16.8941
_Izone_3 |   13.24345   5.624438     2.35   0.019     2.168599
24.31831
_Izone_4 |  -3.979932    5.78657    -0.69   0.492    -15.37403
7.41417
_cons |   1964.247   7284.588     0.27   0.788    -12379.54
16308.04
----------------------------------------------------------------------------
--

. adjust  _Iage_2 _Iage_3 _Iage_4 _Iage_5 , by(period zone) se format(%9.3f)
center

----------------------------------------------------------------------------
--------------------------------------    Dependent variable: salofat
Command: regress
Variables left as is: _Izone_2, _Izone_3, _Izone_4
Covariates set to mean: _Iage_2 = .20557491, _Iage_3 = .29268292, _Iage_4 =
.20092915, _Iage_5 = .07665505
----------------------------------------------------------------------------
--------------------------------------

------------------------------------------------
|       level of contamination
period |  <1 Ci    1-5 Ci  5-15 Ci   >15 Ci
------------+-----------------------------------
spring 2000 |  48.493   54.086   61.736   44.513
| (4.696)  (3.317)  (3.181)  (3.356)
|
fall 2001 |  47.535   53.128   60.778   43.555
| (5.900)  (4.066)  (3.706)  (3.664)
------------------------------------------------
Key:  Linear Prediction
(Standard Error)

The problem is that the mean adjusted on age is not zero has it should have
been in the <1Ci region according to my first table.

Was I wrong in my analysis?
*	If yes, how should have I done?
*	If not, how to explain this result, and how to avoid it?

Bertrand Gagnière
Laboratoire d'épidémiologie
Institut de Radioprotection et de Sûreté Nucléaire
BP 17
92 262 Fontenay aux Roses cedex

Tel: + 33 1 58 35 95 56
Fax: + 33 1 46 57 86 03
<mailto:bertrand.gagniere@irsn.fr>

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```