| censor | stsdas.analysis.statistics | censor |
censor -- Information about the censoring indicator in Survival Analysis.
help censor
Censoring can be categorized in three groups: detection, upper limit, and lower limit. In general, if the data point is detected, the indicator is zero; for an upper limit, the indicator will be less than zero; and for a lower limit, the indicator will be greater than zero. The following table gives possible values and their meaning:
Indicator Independent variable Dependent variable
0 detection detection
1 detection lower limit
-1 detection upper limit
2 lower limit detection
-2 upper limit detection
3 lower limit lower limit
-3 upper limit upper limit
4 upper limit lower limit
-4 lower limit upper limit
For the univariate problems, the indicators that you need are -1, 0, and 1.
If you have more than one data column, you must assign indicators
to each column. Mixing upper and lower limits is not allowed in
any method (i.e., kmestimate and twosampt).
For the correlation and regression problems, the problem is a bit more complicated. The tasks coxhazard, emmethod, and buckleyjames can take censored points only in the dependent variable; the indicators that you can use are -1, 0, 1. Although coxhazard and buckleyjames cannot treat data with mixed censoring, emmethod can (e.g., you can have -1, 0, and 1 in the same data set). These methods all may have multiple independent variables.
The bhkmethod task (i.e., generalized Kendall's tau), generalized Spearman's rho, and schmittbin can treat censoring in the independent variable. The indicators you may use are -4, -3, -2, -1, 0, 1, 2, 3, 4. Although mixed censoring in one variable (e.g., lower and upper limits in the X variable) are not allowed, the cases with lower(or upper) limits in the X variable and upper (or lower) limits in the Y variable can be done.
Note: emmethod is supposed to be able to take confined data points, using a censor indicator of 5. Currently, however, this option does not work and should not be used.