Date:         Thu, 29 Jul 1999 15:43:06 -0400
Reply-To:     "Hamani, Elmaache" <elmaachh@ERE.UMONTREAL.CA>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Hamani, Elmaache" <elmaachh@ERE.UMONTREAL.CA>
Subject:      SAS-Biostatician to help
Content-Type: multipart/alternative;

Hi guys.
 I run  a program on the softwear MLwiN and I have the results. I would like to try it on SAS. But I can.
I will try to explain you as follows the problem:
I am trying  to write the hierarchical model ( 2 levels j is the level 2=Family  and i is level 1=twin), 
with some constrains. My genetic equation is

Yij= m + Cj +R_1j* MZj + R_2j*DZj + R_2ij(DZ)  + e_ij

where Yij response and: Cj , R_1j, R_2j are randon variables of level 2 (or
elements of G in PROC MIXED)
and ( R_2ij ) and ( e_ij )  are randon variables of level 1 (or elements of R in
PROC MIXED)

The two constraints are:

Var(R_1j)=2*Var(R_2j) and Var(R_2j)=Var(R_2ij(DZ) .

By using  PROC MIXED  with LDATA and LIN(q),
I solved  the first constrain and I estimed Var(Cj ), but I can not resolve
the second constrain, that is Var(R_1j)=Var(R_2ij(DZ).

The following is a part of the input:
Family     twin    sexe    Y         MZ     DZ

       1          1        0        13           0       1
       1          2        1        16           0       1
       2          1        0        17           1       0
       2         2         0        17           1       0
       3         1         0        16           1       0
       3         2         0        17           1       0
       4         1         0        14           0       1
       4         2         0        17           0       1
       5         1         0        12           1       0
       5         2         0        14           1       0
       6         2         0        13           0       1
       6         1         1        13           0       1
       7         1         1        22           0       1
       7         2         1        13           0       1
      ;
/*  here my program  */
data con;
   input parm row col1-col3;
   datalines;
1 1  1   0  0
2 2  0   2  0
2 3  0   0  1
;
run;
    PROC MIXED data=un  noclprint  covtest ;
   class family   ;
   model nmis_agr =   / noint;
   random  int dz mz /type=lin(2) ldata=con sub=famille G ;
  run;
/*************************************/
    A  part of   ouput:
                              G Matrix

Effect FAMILY Row    COL1    COL2            COL3

INTERCEPT 1001  1   0.26700391

DZ               1001 2              0.09326706

MZ              1001  3                              0.18653412

      Covariance Parameter Estimates (REML)

Cov Parm Subject    Estimate          StdError     Z      Pr > |Z|

LIN(1)   FAMILLE     0.26700391   0.15804857   1.69   0.0911

LIN(2)   FAMILLE     0.09326706   0.11151294   0.84   0.4029

Residual                0.53656972    0.04241956   12.65  0.0001


               It misses here  DIAG   (DZ) =? for the seconde
relationship:Var(R_2j)=Var(R_2ij(DZ) ).

Thanks  a lot.
best regards.

[text/html]