2
$\begingroup$

I made a library for displaying survival graphs. I am currently struggling with figuring out how to meaningfully display data with 300,000 data points. There doesn't seem to be enough pixels in the area of the graph.

Would it makes sense to do something like "higher tickmarks for more people that fall on the same day"?

$\endgroup$
  • 1
    $\begingroup$ Are you saying that on timepoints when only a few events occur the survival curve does not go down, because the relatively small drop is not made visible due to a lack of pixels? Aren't you able to increase the dots (i.e. pixels) per inch? Alternatively you could try using vector graphics, where zooming in should show small drops too. Further, a Kaplan-Meier plot already contains a measure of how many events occur at a timepoint: the size of the drop (i.e. the more events have occurred, the further you go down). $\endgroup$ – IWS Apr 11 '17 at 14:42
2
$\begingroup$

Generally speaking, tick marks on a Kaplan-Meier plot are to indicate censored observations, not more events at the same time. However, it is common that you have the option to turn the ticks off (i.e., not display them). Instead of having higher ticks for more censored observations at a given time, I would just have them not displayed by default when N is large. On the other hand, asymptotic confidence intervals should be very reliable when you have very large N. Having the curve with the confidence interval strikes me as conveying the useful information more effectively than trying to represent the censorings. Thus, when N is large I would default to having the curve lay on top of a shaded background representing the confidence band. Here is an example, copied from here (I might not make it red, though):

enter image description here

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.