I am reviewing an article, so I can't be too specific, but the authors are doing a survival analysis of outcomes of stroke and one of the independent variables is the time that the person had a stroke - one hypothesis is that treatment got better over time.
The authors have divided "time entering the trial" into 4 groups, each of three years (e.g. 2000-2002, 2003-2006 etc) and plotted those curves. However, since there is also right censoring (for the end of the study) the curves have different lengths.
This strikes me as less than ideal, but I am not if anything is better. Is there something?