Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • I am confused about the confidence interval of survival analysis between two methods

    If I analyze a set of data use syntax ”ci means variate” and I get a result like this

    ci means y

    Variable | Obs Mean Std. Err. [95% Conf. Interval]
    -------------+---------------------------------------------------------------
    y | 47 4.929787 .4511155 4.021739 5.837836



    If I use survival analysis syntax “stci” and I get thiswho can tell me which one I can count on

    stci

    failure _d: 1 (meaning all fail)
    analysis time _t: y

    | no. of
    | subjects 50% Std. Err. [95% Conf. Interval]
    -------------+-------------------------------------------------------------
    total | 47 5.6 .3954528 1.4 6.3



    Thank you very much

  • #2
    It appears that you have survival data and you want the mean duration. A key aspect of survival data is that you typically have censored observations; at some point you have to stop collecting data even if not everybody in the sample has experienced the event yet. So for some persons you know that the haven't had the event for say the first 3 months but you don't know when they will get it. The way this information is stored is in two variables: the duration, and the censoring variable. The duration variable contains for those persons who experienced the event the time till event and for those person who did not experience the event the time till you stop measuring. The censoring variable tells you which person experienced the event and which person did not. So the duration variable can mean two different things, and the censoring variable tells you what meaning is appropriate for what person. As a consequence just trying to summarize the duration variable in isolation mixes up these two times and is a bad idea. So in that case ci means is a bad idea.

    Also note that stci does not give you the mean but the median. That makes sense, as in order to compute the mean you need all the times till the event and you don't have those for the censored observations. For the median you only need the sort order. If more than half the observations experienced the event, then you know that the censored observations took longer than that and that is all you need to compute the median.
    ---------------------------------
    Maarten L. Buis
    University of Konstanz
    Department of history and sociology
    box 40
    78457 Konstanz
    Germany
    http://www.maartenbuis.nl
    ---------------------------------

    Comment


    • #3
      That's very kind of you,i seems that i should use stci syntax to get 95%CI,especially in survival analysis. but i dont no what kind of situation the ci means should selected?

      Comment

      Working...
      X