hello Steve, things are becoming a bit clearer to me, thanks very much for the straight tips. Looking at the graph I created (and attached), I was not entirely surprised that genotype*sweek is not significant. Or maybe I should have been ...? The oddsratios for the 4 genotypes in the common slope model are not the same at each time point or, in other words, not the same no matter the value I enter for the <at= option> (as I understand you), but they are exactly the same for the same sweek- intervals I'd like to look at in order to find out, from when to when the risk of score1=1 significantly changes in each genotype (i.e. the same for any value I enter for the <unit= option>; e.g. ORs (at sweek= -2 unit sweek=4) genotype sweek genotype sweek estimate DF 95% Confidence limits ... ... ... 1 -2 1 2 0.615 604 0.469 0.808 2 -2 2 2 0.618 604 0.524 0.730 3 -2 3 2 0.718 604 0.561 0.919 4 -2 4 2 0.724 604 0.565 0.927 are the same as for the following 4 week intervals (at sweek=2 unit sweek=4; and so forth) and stays significant. It's the same if I try to look at 6-week-intervals ... The estimate of the common slope is -0.0909, which is not much, isn't it? From the chart it doesn't look as if it would be so small. I am also not sure how to interpret the quadratic sweek term being signficant whereas there is equality of quadratic regressions (sweek*sweek*genotype is not significant); slope for sweek= -0.199 and for sweek*sweek= 0.00474 if both are included). Maybe because after a decrease there is some degree of a small increase in the risk for score1=1 as sweek increases... The example you mentioned uses the type 1 tests of fixed effects, which I did not. Edit: The analysis by genotype has resulted in similar estimates for the slope sweek in the 4 genotypes, but not the same: (1) -0.1200, (2) -0.1176, (3) -0,1086, (4) -0,08795 The corresponding oddsratios are 0.619, 0.625, 0.648 and 0.703 and stay the same for all 4 week intervals. I can really not believe that the risk for score1=1 (becoming lame) decreases over time intervals in a constant manner in each of the 4 genotypes.
... View more