w******8
发帖数: 59 | 1 I am a new biostatistician and I am so confused when the medical doctor
wants me to see which variables can
predict event-free survival. Less than 60 patients were on trial and only 31
patients had events at the time of
analysis. But we are looking at about 300 variables, of which many are
longitudinal (repeated measures) and
are correlated. As I read in some book, the number of variables that can
enter a model is about 4th root of
the number of events. So technically I can only enter about 2 to | b*******r
发帖数: 152 | 2 try starting off with the single-var model to see if the signs of the parms
are right or not. |
|
