On Thu, 4 Nov 2004, K Fernandes wrote:

> Suppose I have a dataset called Summary1 which is this: > > Rand Sim_No Cen_Num x m y > 1 1 1 0 1 0 > 1 1 1 1 4 0 > 1 1 2 0 3 1 > 1 1 2 1 4 2 > 1 1 3 0 5 1 > 1 1 3 1 4 2 > 1 2 1 0 2 1 > 1 2 1 1 1 1 > 1 2 2 0 1 0 > 1 2 2 1 3 0 > 1 2 3 0 3 3 > 1 2 3 1 3 2 > > (My actual dataset has more than one value in the Rand column (i.e. 1 to > 5) and more than two values in the Sim_No column (i.e. 1-3). >

<snip>

> > For 1), I tried this: > proc logistic data = Summary1; > model y/m = x ; > ods output ParameterEstimates = BetaEsts; > by Rand Sim_No; > run; >

> However, the estimates I get for 1) does not correspond to estimates I > get in R, so I think there is a problem with my statements. Can anyone > spot it?

Kim

LOGISTIC applies 'effects' coding by default to the coefficients. Even though x is 0/1 you might want to try it with the CLASS statement with the 'glm' (or 'ref') option for param= as indicated below. Also, the sorted order of the data (i.e. levels of x within the ByVar levels) will determine the sign of the coefficient of x.

try....

PROC SORT DATA=Summary1; BY Rand Sim_No x;

proc logistic data = Summary1; CLASS x / param=glm order=data; model y/m = x Cen_Num; by Rand Sim_No; run;

or......

PROC SORT DATA=Summary1; BY Rand Sim_No descending x;

proc logistic data = Summary1; CLASS x / param = glm order=data; model y/m = x Cen_Num; by Rand Sim_No; run;

Robin High Univ. of Oregon