Skip to content
Prev 58738 / 398502 Next

survSplit: further exploration and related topics

To Danardonos concern of splitting time for records with delayed entry:

This can fairly easily be accomodated, by simply splitting time in small
intervals of time since entry into the study, and then compute the value
of the other timescales for each of these e.g.:

current.age <- time.from.entry + age.at.entry

but the cut on the other timescales will not be exactly where you may
want 
them to be, but this should be of minor importance in real life.

This approach will also make it clearer what really is going on.

The effect of each of the timescales can then be modelled using the
usual
regression tools available in R.

David Clayton har written an R-function that does it correctly, you can
find it in:
http://biostat.ku.dk/~bxc/Lexis/ along with its .Rd file.

It is also included a Lexis-package which is a first shot at an
epidemiology package 
for R available at http://biostat.ku.dk/~bxc/SPE/library/, but built
under 1.9.
I recently tried to build it under 2.0, but it crashed for me and I was
advised to
wait till 2.1 before I had another go at it.

A little further exploration of what goes on in survSplit gives:
+                     event="status",episode="i")
id expand start time status          x
11  k     30     0    5      0 Maintained
34  k     31     5   10      0 Maintained
57  k     32    10   50      0 Maintained
80  k     33    50  161      0 Maintained
+                      event="status",episode="i",id="id2")
aml4[aml4$id=="k",c("id","expand","start","entry","time","status","x")]
    id expand start entry time status          x
30   k     30     0     0    5      0 Maintained
31   k     31     5     0    9      0 Maintained
94   k     31     5     9   10      0 Maintained
32   k     32    10     0    9      0 Maintained
95   k     32    10     9   12      0 Maintained
158  k     32    10    12   40      0 Maintained
221  k     32    10    40   50      0 Maintained
33   k     33    50     0    9      0 Maintained
96   k     33    50     9   12      0 Maintained
159  k     33    50    12   40      0 Maintained
222  k     33    50    40  161      0 Maintained
+                      event="status",episode="i",id="id2")
id expand start time status          x
30   k     30     0    5      0 Maintained
31   k     31     5    9      0 Maintained
94   k     31     9   10      0 Maintained
95   k     32     9   12      0 Maintained
158  k     32    12   40      0 Maintained
221  k     32    40   50      0 Maintained
96   k     33     9   12      0 Maintained
159  k     33    12   40      0 Maintained
222  k     33    40  161      0 Maintained
It appears that the intention has been to support counting process
input, 
but not quite succeeded.

Bendix Carstensen
----------------------
Bendix Carstensen
Senior Statistician
Steno Diabetes Center
Niels Steensens Vej 2
DK-2820 Gentofte
Denmark
tel: +45 44 43 87 38
mob: +45 30 75 87 38
fax: +45 44 43 07 06
bxc at steno.dk
www.biostat.ku.dk/~bxc
----------------------