I have question for those more familiar with statistics. Iím a lab geek, so I have functional knowledge of these stats, but when things start getting into algorithms and such Iím going to need some help.

My statistics question concerns the SPSS output for Cox regression survival analysis and how it compares to R output.

Below will be a picture of the example data.

In the KM graph is looking at survival rates given a mutation or not. And deaths are occurring at the actual time. This part looks good as far as Iím concerned. In the Cox graph, age has been added as a covariant. I know that the Cox regression is representation of PREDICTED survival and not actual survival; however the step downs in the two lines perfectly match each other. It makes it look as if deaths are happening at the exact same time in both groups.

Apparently, statisticians using R donít see this perfect mirroring of the two lines, and they are very insistent that SPSS is incorrect. Iím pretty certain the SPSS analysis is right and there are countless examples of it looking like this online. It also looks like SAS outputs the same thing. And Iím not familiar with R, but when I asked around, depending on what package is being used, sometimes the output looks mirrored like that and sometimes it looks more what youíd expect from a KM.

I just need some information, and preferably something I can cite like a white paper that explains how and why the software chooses to make the stepped pattern basically identical between the two groups. And why it might be different than what is seen with certain R packages. What is the actual difference between these two when they are apparently doing the same test.

This might be a very simple straight forward answer for people more familiar with the algorithms. Iíve looked through the user manual and community groups online for the answer and have had little luck.

Thank you to anyone that can address this in some way that is understandable to a novice like me.