This is how I think about possible ways to use stset to analyse a
longitudinal study. For further discussion of these issues please see
Part D of Essential Medical Statistics by Kirkwood and Sterne.
In the following, variable exitdate contains each subject's date of
exit from the study. Variable examdate contains subjects' date first
examined, and variable dob contains subjects' date of birth. All these
variables are integer date variables counted in days since 1/1/1960:
type help dates for details. Variable y contains the follow up time in
years, and d contains the outcome.
1. Simplest possible use of stset, when time in study (y) has already
been derived:
stset y, failure(d)
2. Equivalent, using dates of entry and exit directly
stset exitdate, failure(d) origin(examdate) scale(365.25)
In methods 1 and 2, time 0 (the origin) is the time of entry to the
study. These methods are appropriate for studies following individuals
from a particular event such as diagnosis of disease or recruitment to
a randomised trial.
3. Using date of birth, entry date and exit date to count time in
years of age:
stset exitdate, failure(d) enter(examdate) scale(365.25) origin(dob)
Method 3 is likely to be appropriate for long term studies in which
subjects' age, rather than the time that they have been in the study,
is likely to be the most important determinant of their risk of
disease. This method implies that subjects enter the study at
different times (delayed entry).
4. Counting from a fixed point in time.
When you type:
stset exitdate, failure(d) enter(examdate) scale(365.25)
time is counted from a fixed point of 1/1/1960, with entry to the
study at each subjects examdate. This is unlikely to be sensible, but
a small modification would allow us to count time from a fixed date.
This would be useful if we were following subjects from the time of a
particular incident such as an environmental disaster. If the date of
interest was 17 June 1966, we could type:
gen origdate=mdy(6,17,1966)
stset exitdate, failure(d) enter(examdate) origin(origdate) scale
(365.25)
Method 4 is rarely used, but would be useful in a study that followed
subjects from the time of a particular incident, but with delayed
entry at different times after the incident.
I hope this is helpful.
Best wishes
Jonathan Sterne
> Date: Wed, 22 Mar 2006 12:46:56 +0000
> From: Sue Chinn <[email protected]>
> Subject: st: left-truncation of entry in survival analysis
>
> Dear Statalist readers,
>
> Reports of survival analysis which use age as the time scale rather
> than
> time-on-study often 'adjust for delayed entry'. In Stata this is
> achieved by:
>
> stset age, fail(died) enter(ageatentry)
>
> (see recent e-mail from Dawn Teele, or reply to st: streg from
> [email protected] on 19th September 2002.)
>
> However, a model fitted with the above stset gives exactly the same
> answer
> as one with
>
> stset timeonstudy, fail(died)
>
> provided timeonstudy=age-ageatentry (as it normally would, but might
> not
> exactly depending how variables were calculated) and the models are
> exactly
> the same. In the second model it is usual to adjust or stratify on
> age,
> while in the first it isn't as age is taken into account, supposedly,
> so
> researchers may not have realised the equivalence.
>
> So, am I missing something, or are advocates of the first model
> deluding
> themselves? Can left truncation be ignored with age as the
> timescale?
>
> Thanks
>
> Sue
>
> Sue Chinn
> Professor of Medical Statistics
> Division of Asthma, Allergy and Lung Biology
> King's College London
> 5th Floor Capital House
> 42 Weston Street
> London SE1 3QD
>
> tel no. 020 7848 6607
> fax no. 020 7848 6605
>
>
_______________________________________
[email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
