I have been using the Cox proportional model option in SPSS to estimate relative risk between two groups (which appears as the Exp B statistic in the output). But the estimates I get don't seem to reflect the survival plots. I tried to do some calculations by hand using the life table outputs. These estimates are different from the SPSS outputs. Has anyone experienced a difference between manually calculated relative risk and SPSS computed RR, and if so how did you reconcile the difference?
Hi,
if I I have understood correctly you problem you have noted that B values do not seems to reflect the KM curves proposed by SPSS. If it is correct, I can suppose that the KM curves you have obrtained have all very similar shape. This is because, if I remember well, the Cox analysis propose extimated KM curves, which are different form the curves you may obtain by unadjusted KM analysis. I hope this could be of help for you.
Hi,
if I I have understood correctly you problem you have noted that B values do not seems to reflect the KM curves proposed by SPSS. If it is correct, I can suppose that the KM curves you have obrtained have all very similar shape. This is because, if I remember well, the Cox analysis propose extimated KM curves, which are different form the curves you may obtain by unadjusted KM analysis. I hope this could be of help for you.
Yes, If you perform survival analyses using the KM test with the SPSS you will obtain the curves but the results are not adjusted for confounding factors. With Cox regression you can adjust for confounding factors and obtaint relative risk values more accurate and the estimated survival curves after adjustment.
Yes the Hazard Ratio of Cox Regression is likely to differ from the Life table estimates as it give the Survival function and Cox delays with Hazard function with its baseline hazard which cannot be got from life table method.
Simple answer: either you believe on your calculation or SPSS. But my suggestion is trust one of them.If you are statistician, then find the way, if not try to learn how to interpret what you got
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS