I am disecting my Kaplan-Meier curves, of four different phenotypes of C. elegans, into first 8 days and last 20 days, in order to analyse in which period there is significant difference in the moratlity rate. I am unsure as to whether inculding/excluding the living fractions, relevant to the time-frame being examined, as censored data (censored data is representative of fraction(s) living beyond the time of the event, e.g. end of experiment or after 8 days) would confound the statiscal results.
It would be a big mistake (biased results) not to include censored data in your KM curve and log-rank test.
Thank you for your answer Renaud. I suspected that to be the case, but when comparing the data from the last 20 days, with included censored data, the starting population (at day 8) and by extension the end result, due to different mortality rates in the previous 7 days, will be different. Isn't that a problem? And if it is, is there any way around it?
Well, that's real life data... If I correctly understand what you mean, we discussed this kind of problem in the attached paper. See table 2.
You still get to understand the survival probabilities at various points. As Renaud indicates, that's real life. Different tests are affected by censoring patterns, so you might want to do several , checking and to see which are most and least affected by censoring at different points in the risk period. I think, personally, that you report your findings and describe the censoring pattern. That's what is most helpful to readers. imo.
Best wishes
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics