SURVIVAL <data>,<time-variable>,<censoring-variable>,<output line>
Programmed by K.Huuhko (last modification 17.10.2002)
Documentation R.Sund 11/2004

SURVIVAL-module is a package of techniques for non-parametric survival
analysis. The basic methods include Kaplan-Meier estimator, life table
estimator and three test statistics for similarity of survival curves:
log-rank, Wilcoxon and Tarone-Ware.

Time-variable is should be a variable containing time to event or time
to censoring, and censoring variable indicates whether the observation
is censored (value 0) or not (any other value). Without
METHOD-specification the module calculates the Kaplan-Meier estimator
and with specification METHOD=LT the life table estimator is calculated.
The results are written to the data file PLc0.SVO and plotting schema is
printed to the edit field.

For life table estimator specification INTERVAL can be used to define
the width of time interval used in calculation. The life table is also
printed to the edit field, if it contains fewer intervals than given
with the specification LIMIT.

With the specification CLASS the name of classification variable can be
given, where classification variable gets equivalent value for all
observations belonging to some particular group. In that case analyses
are done separately for each group (results saved for files
PLc0,PLc1,...) and also test statistics for similarity of survival
curves are calculated.

Using the specification CLASS=TREE a survival tree is calculated. The
variables used in splitting must be marked with mask X. The calculation
can be affected by using the specifications PENALTY, OBSLIMIT and TEST.
PENALTY is a "nuisance" parameter which can get any value between 2 and
4. Four approximately corresponds to p-value 0.05, and smaller values
allow also "less significant" splits. OBSLIMIT gives the minimum number
of observations needed in the end nodes, i.e. how many observations is
needed that a further split can be made for the node. The specification
TEST determines the test statistics used in the tree construction:
1=log-rank, 2=Wilcoxon or 3=Tarone-Ware. After initial tree construction
a specification NODES=<wanted number of terminal nodes> must be given,
and the information concerning the binary survival tree is printed to
the edit field. This information starts from the node one (whole data)
and describes each split growing the tree to the left until the first
end-node is reached (survival curve is estimated for each end-node and
saved to the data file T<number of end-node>.SVO) and then climbing back
up to the nearest right-split and so on. Typically a tree resulting in 2
to 8 end-nodes is sufficient.

Cautions:
The module can not handle MISSING(/negative?/wrong type???)-values.

With Kaplan-Meier and life table estimators the degrees of freedom (and
therefore also p-values) are calculated incorrectly in cases of more
than two classes.


.......................................................
SAVE MYEL  / Esimerkki

DATA MYEL
AIKA CEN SEX   NIT HEMO AGE CALC
   1   1   0 2.218  9.4  67   10
   1   1   0 1.940 12.0  38   18
   2   1   0 1.519  9.8  81   15
   2   1   0 1.748 11.3  75   12
   2   1   0 1.301  5.1  57    9
   3   1   1 1.544  6.7  46   10
   5   1   1 2.236 10.1  50    9
   5   1   0 1.681  6.5  74    9
   6   1   0 1.362  9.0  77    8
   6   1   1 2.114 10.2  70    8
   6   1   0 1.114  9.7  60   10
   6   1   1 1.415 10.4  67    8
   7   1   0 1.978  9.5  48   10
   7   1   1 1.041  5.1  61   10
   7   1   1 1.176 11.4  53   13
   9   1   0 1.724  8.2  55   12
  11   1   0 1.114 14.0  61   10
  11   1   0 1.230 12.0  43    9
  11   1   0 1.301 13.2  65   10
  11   1   0 1.508  7.5  70   12
  11   1   1 1.079  9.6  51    9
  13   1   1 0.778  5.5  60   10
  14   1   0 1.398 14.6  66   10
  15   1   0 1.602 10.6  70   11
  16   1   0 1.342  9.0  48   10
  16   1   1 1.322  8.8  62   10
  17   1   0 1.230 10.0  53    9
  17   1   0 1.591 11.2  68   10
  18   1   1 1.447  7.5  65    8
  19   1   0 1.079 14.4  51   15
  19   1   1 1.255  7.5  60    9
  24   1   1 1.301 14.6  56    9
  25   1   0 1.000 12.4  67   10
  26   1   1 1.230 11.2  49   11
  32   1   0 1.322 10.6  46    9
  35   1   0 1.114  7.0  48   10
  37   1   0 1.602 11.0  63    9
  41   1   0 1.000 10.2  69   10
  42   1   1 1.146  5.0  70    9
  51   1   0 1.568  7.7  74   13
  52   1   1 1.000 10.1  60   10
  54   1   0 1.255  9.0  49   10
  58   1   1 1.204 12.1  42   10
  66   1   0 1.447  6.6  59    9
  67   1   0 1.322 12.8  52   10
  88   1   1 1.176 10.6  47    9
  89   1   0 1.322 14.0  63    9
  92   1   1 1.431 11.0  58   11
   4   0   0 1.945 10.2  59   10
   4   0   1 1.924 10.0  49   13
   7   0   1 1.114 12.4  48   10
   7   0   0 1.532 10.2  81   11
   8   0   1 1.079  9.9  57    8
  12   0   1 1.146 11.6  46    7
  11   0   0 1.613 14.0  60    9
  12   0   1 1.398  8.8  66    9
  13   0   1 1.663  4.9  71    9
  16   0   0 1.146 13.0  55    9
  19   0   1 1.322 13.0  59   10
  19   0   1 1.322 10.8  69   10
  28   0   1 1.230  7.3  82    9
  41   0   0 1.756 12.8  72    9
  53   0   0 1.114 12.0  66   11
  57   0   0 1.255 12.5  66   11
  77   0   0 1.079 14.0  60   12

.......................................................
SURVIVAL MYEL,AIKA,CEN,END+2
.......................................................
METHOD=LT INTERVAL=10
SURVIVAL MYEL,AIKA,CEN,END+2
.......................................................
CLASS=SEX
SURVIVAL MYEL,AIKA,CEN,END+2
.......................................................
CLASS=TREE VARS=SEX(X),NIT(X),HEMO(X),AGE(X),CALC(X)
SURVIVAL MYEL,AIKA,CEN,END+2
.......................................................
NODES=3 CLASS=TREE VARS=NIT(X),HEMO(X),AGE(X),CALC(X)
SURVIVAL MYEL,AIKA,CEN,END+2
.......................................................