Output | Predictive Modeling | Results (Proportional Hazards Regression)

Results (Proportional Hazards Regression)
The Results tab is shown below:
The Results tab contains the following elements:
 •
This plot shows the survival rate across the range of observed event time for each individual or sample. The y-axis indicates the survival rate and the x-axis indicates the event time.
Note: If you specify a Group Variable on the General tab, each group is represented by a different color.
 •
This plot shows the distribution of survival rate for each group in box plot.
The y-axis indicates the estimated survival rate for the reference time, 7.5 in this example. You can set the reference time using the Reference Time for Comparing Survival Curves option on the Analysis tab of the process dialog. If you leave this parameter empty, the median among all samples is applied.
The x-axis in the plot indicates the corresponding groups (NIC 15 and Placebo, in this example). The horizontal line (at 0.9) indicates the overall average. The left panel of the plot shows the result of comparison between groups using Hsu's MCB method. When the two circles are disjoint, the corresponding average survival rates are significantly different (in this example they are not).
You can apply other multiple comparison methods by clicking on the red triangle (circled in blue) and selecting one of the Compare Means options (as shown below).
There are certain modeling settings and results shown above the plots. The settings for predictor reduction and analysis are listed in lines one and two respectively. The selected predictors are listed on line three. The training set criteria (in this case, Harrell's C Statistic1, is listed in line four. The reference survival time applied to the box plot on the right is shown in line five.
In this example, the points on the box plot are the survival rates of event time at 7.5 on the left survival plot.

Harrell’s C Statistic is a concordance measurement between the observed event times and the area under estimated survival curves (AUC) among samples. Samples with larger event times tend to have larger corresponding AUCs. In this case, their associated Harrell's C statistics are larger. The Harrell's C Statistic can be interpreted as the proportion of concordant pairs among the suitable pair-wide comparisons across all samples.