[R] RES: survival

[Thomas Lumley]

On Wed, 8 Mar 2006, Paulo Brando wrote:
>
>>  summary(model.fit) # just one species from one treatment shown below
>
> Call: survfit(formula = Surv(time, censo) ~ treatment + species, data =
> wsuv)
>
>                treatment=0, species=1
> time n.risk n.event survival std.err lower 95% CI upper 95% CI
>    1  15440     386    0.975 0.00126        0.973        0.977
>    2  15054     336    0.953 0.00170        0.950        0.957
>    3  14668     302    0.934 0.00200        0.930        0.938
>    4  14282     296    0.914 0.00226        0.910        0.919
>    5  13896     281    0.896 0.00247        0.891        0.901
>    6  13510     264    0.878 0.00264        0.873        0.883
>    7  13124     251    0.861 0.00280        0.856        0.867
>    8  12738     232    0.846 0.00293        0.840        0.852
>    9  12352     216    0.831 0.00305        0.825        0.837
>   10  11966     206    0.817 0.00315        0.811        0.823
>   11  11580     190    0.803 0.00325        0.797        0.810
>   12  11194     179    0.790 0.00333        0.784        0.797
>   13  10808     167    0.778 0.00341        0.772        0.785
>   14  10422     167    0.766 0.00349        0.759        0.773
>   15  10036     145    0.755 0.00356        0.748        0.762
>   16   9650     142    0.744 0.00363        0.737        0.751
>   17   9264     135    0.733 0.00369        0.726        0.740
>   18   8878     122    0.723 0.00375        0.715        0.730
>   19   8492      99    0.714 0.00380        0.707        0.722
>   20   8106      84    0.707 0.00385        0.699        0.714
>   21   7720      68    0.701 0.00389        0.693        0.708
>   22   7334      66    0.694 0.00393        0.687        0.702
>   23   6948      51    0.689 0.00397        0.681        0.697
>   24   6562      40    0.685 0.00400        0.677        0.693
>   25   6176      38    0.681 0.00403        0.673        0.689
>   26   5790      37    0.676 0.00407        0.669        0.684
>   27   5404      33    0.672 0.00411        0.664        0.680
>   28   5018      31    0.668 0.00415        0.660        0.676
>   29   4632      26    0.664 0.00419        0.656        0.673
>   30   4246      22    0.661 0.00423        0.653        0.669
>   31   3860      15    0.658 0.00427        0.650        0.667
>   32   3474      14    0.656 0.00431        0.647        0.664
>   33   3088      14    0.653 0.00436        0.644        0.661
>   34   2702      13    0.650 0.00443        0.641        0.658
>   35   2316      12    0.646 0.00451        0.638        0.655
>   36   1930      11    0.643 0.00462        0.634        0.652
>   37   1544      12    0.638 0.00480        0.628        0.647
>   38   1158      10    0.632 0.00507        0.622        0.642
>   39    772       9    0.625 0.00557        0.614        0.636
>   40    386       8    0.612 0.00709        0.598        0.626
>
> I don't get why with 8 leaves remaining (out of 384), the survival is
> about 0.6???
>

It looks as though the majority of your leaves are censored, especially at 
later time points. At each of your 40 time points about 1-2% of the leaves 
under observation die, so the survival curve should end up somewhere 
between 0.98^40 =0.45 and 0.99^40=0.67, and it does.

 	-thomas