I am currently working on a Cox regression in SPSS (Complex Samples). The cox analysis has age as time scale and calendar year as time-dependent variables, while treatment (y/n) is a factor variable.
My problem is that I don't know how to confirm the proportional hazards (PH) assumption when the time scale is set to age. I've tried a stratified analysis with a -log-log plot. There is an intersection at the very start of the plot. Can I go forth with this and assume that this has no significance on the PH assumption or do I need to make a stratified analysis for my "final" analysis?

Also, when I use the "Test for proportional hazards" option instead of the graphical -log-log interpretation in SPSS I get the following output:
Overall test of proportional hazards:
df1 3.00 df2 15975 Wald F .322 Sig. .809

Now, I thought the PH assumption was fulfilled when the Sig (P-value) was >0.05, but now I don't know.

I hope there are some smart biostatisticians out there. I am kind of a rookie and I've tried googling around but without luck.

Best regards,