Survival Analysis time feature
Survival Analysis time feature
When I use Survival Analysis, the chart comes out showing the time elapsed as running from 1 to 10. As my times - as shown in the data - run from 40 to 180 days - this isn't helpful. I think that the user should see the time they have specified, with a median of that time, not a notional '1 - 10'.
- Attachments
-
- Transfers.omv
- (14.08 KiB) Downloaded 103 times
-
- Transfers.csv
- (179 Bytes) Downloaded 99 times
Re: Survival Analysis time feature
try changing the variable to be continuous.
jonathon
jonathon
Re: Survival Analysis time feature
Thanks Jonathon. I'll write it into my new editions as a must-do as part of the process.
Re: Survival Analysis time feature
we should probably improve it, so it handles that though.
jonathon
jonathon
Re: Survival Analysis time feature
If you could let me know if/when changes are under way, I would be grateful. Although I'm well into work on the second editions, I can wait if there are imminent changes. www.vorpress.com
Re: Survival Analysis time feature
A similar development issue: 'Censored' often comes up as the 'Event level' and needs changing. E.g. in the recent file, 'transfer' was the event and 'Censored' was the alternative (this is because R works on an alphabetical default). Is it possible for the program to spot that this has come up, and choose the alternative term as the event? or will it be messed up by different spellings of 'Censored' or other usage?
Re: Survival Analysis time feature
hi cole,
sorry for the delay. i've pushed out deathwatch 1.1.2 which will automatically handle the situation where the 'elapsed' variable isn't continuous.
i think we're reluctant to 'auto-detect' the correct 'event level' ... if we guess it right some times, but not other times, people will blame us when we get it wrong ... where as if we just don't do anything, we can avoid blame altogether
cheers
jonathon
sorry for the delay. i've pushed out deathwatch 1.1.2 which will automatically handle the situation where the 'elapsed' variable isn't continuous.
i think we're reluctant to 'auto-detect' the correct 'event level' ... if we guess it right some times, but not other times, people will blame us when we get it wrong ... where as if we just don't do anything, we can avoid blame altogether
cheers
jonathon
Re: Survival Analysis time feature
Thank you, Jonathon, fair point.