Prism offers lots of choices when graphing survival data. Most of the choices are present in both the Welcome dialog and the Format Graph dialog, others are only present in the Format Graph dialog. Many are also present in the Parameters dialog for the survival analysis.
These choices are straightforward matters of taste:
•Plot survival or deaths? The former, used more commonly, starts at 100% and goes down. The latter starts at 0% and goes up.
•Plot fractions or percents? This is simply a matter of preference. If in doubt, choose to plot percentages.
You can change the appearance of the graph either in the Format Graph dialog or the Parameters dialog of nonlinear regression. If the graph only plots one survival analysis, the two are equivalent. If you plot two or more survival analyses on one graph, then the Format Graph dialog doesn't always work and it is best to make the changes on all the Parameters dialogs.
As shown above, survival curves are usually plotted as staircases. Each death is shown as a drop in survival.
In the left panel, the data are plotted as a tick symbol. These symbols at the time of death are lost within the vertical part of the staircase. You see the ticks clearly at the times when a subject's data was censored. The example has two censored subjects in the treated group between 100 and 150 days.
The graph on the right plots the data as circles, so you see each subject plotted.
Showing error bars or error envelopes make survival graphs more informative, but also more cluttered. The graph on the left above shows staircase error envelopes that enclose the 95% confidence interval for survival. This shows the actual survival data very well, as a staircase, but it is cluttered. The graph on the right shows error bars that show the standard error of the percent survival. To prevent the error bars from being superimposed on the staircase curve, the points are connected by regular lines rather than by staircases.