jason_zhu@my-dejanews.com wrote: > > Hi all: > > I have a data set about patients’ locomotion. The dependent var is the > level of locomotion, with 1 as very hard to move and 10 as no problem > to move, and 2 to 9 in between, answered by patients. The independent > vars are temperature at admission, age, type of surgeries etc. We did 3 > waves of survey, while during the period of time, some 5% patients > died. The question is how to deal with those died? Do we have a model > to fit them in? Should we assign their locomotion level as 0 (certainly > worse than 1) or as missing? I do not want to toss them away. > > Many thanks! > > Jason > > --== Sent via Deja.com http://www.deja.com/ ==-- > ---Share what you know. Learn what you don't.---

Jason,

Zero is probably not a very good choice: How about those who are still alive but cannot move at all? In the future, you may want to account for those, too. Missing would normally have the meaning 'data not available'. I am afraid this leaves us with the choice of -1... If I were dead, I would most likely consider my level of locomotion quite negative.

Paul M. Dorfman, locomotion level 10, Jacksonville, FL