[R] Survival analysis

cpoiw@rt m@iii@g oii chemo@org@uk cpoiw@rt m@iii@g oii chemo@org@uk
Sun Apr 26 13:20:54 CEST 2020

On 2020-04-26 10:48, Medic wrote:
> Very grateful for the all comments!
> My data contains:
> • left censored
> • right censored
> • events
> (interval censored does not contain!)
> (P.S. I understood, that the code with "type = 'left'" is not
> suitable, because is ONLY for left-censored.)
> I wanted to get the appropriate code for my so mix data mainly in
> order to get a graphical display (graphic aspect) of such model.

I know why you want the graphic.  We all want the graphic! But - the 
graphic is the end result of the model.  I'm worried that you know what 
you WANT the graphic to show and are trying to manipulate the model to 
show it.

I assume you aren't able to share the detail of what you are doing in 
terms of the data?  But can you share the concept?  By that I mean - 
what is the event, what does the left censoring mean etc.  Most studies 
show data from t=0 i.e. day of recruitment / randomisation.  There may 
be logic not to doing that,  for instance - I am involved in a study 
that recruits at t=0, but doesn't randomise until t=56 (if the event has 
occurred randomisation wouldn't happen in this study). It is possible 
that in that 56 days an event can occur (although relatively rare).  
(The stats for that study are not my problem so I don't know what their 
plan is - but it may require left truncating, either in the model - or 
simply only including patients who made it to randomisation in the 
analysis, and possibly taking the event analysis as time from 
randomisation, but that should all be in the stats plan so that the 
numbers can't be played later)

Provide us 10 lines of (interesting) data and talk us through what it 
means.  We don't need to know what "treatment" they had. We don't event 
need to know what the event was other than that the event can only 
happen once (death), or could happen more than once to the same patient 
(heart attack*).  *if that is the case tell us how a second event is 
shown in the data.

This is not a stats help forum, but if its simply sorting data 
structure, that is probably on topic.  If its is this the right test - 
thats off topic.

Have you already collected the data and are now doing analysis? or are 
you trying to structure the data in advance of analysis?

More information about the R-help mailing list