Great that you left a download link!

Best regards,
Reinhard

I noticed some difficulties with the regular coxph() pspline and after some debugging I’ve changed the pastebin (se prev. comment) to a version that works with the following test code:


hmohiv<-read.table("http://www.ats.ucla.edu/stat/R/examples/asa/hmohiv.csv&quot;, sep=",", header = TRUE)

fit <- cph(surv~ rcs(age, 5), data=hmohiv, x=T, y=T)
par(xaxs="i", yaxs="i")
plotHR(fit, terms="age", bty="l", xlim=c(25, 55), y.ticks=c(1,2,3), ylog=F, rug="ticks")


I’ve also changed the density polygon so that it doesn’t need a second plot but is created in the main canvas – this way you don’t need the dens.pos – the density is always positioned at the x-axis.

Another thing I changed was the ability to add exp() to the spline in case someone prefers that format on the yscale.

I worked a little more with the plotHR adapting it to my needs. You can find the latest version here: http://pastebin.com/v0LZje5F

What I’ve done is to add the possibility of plotting mutliple models in the same plot. As I’ve posted previously I’m struggling with the finding out how to do time dependent coefficients in R and right now I’m chopping up the time and plotting the graphs on top of each-other. It’s not a dream solution but it does the job OK.

I’ve also realized what the waist of the cph() cox regression represents – it is the reference that all the others relate to. If you do a:

abline(h=0, col=rgb(0.5, 0.5, 0.5, 0.5), lty=2)
abline(v=median(data[, term.label]), col=rgb(0.7, 0.7, 0.7. 0.5), lty=2)

you get a cross through the waist. I guess it’s not so strange but now I’m a little confused how the regular coxph() does this without a reference. After all cph() should only be a wrapper around coxph().

Thanks once more for your excellent software

Regards
Max

1. The graphical user interface is very intuitive and understandable
2. The backup is encrypted in the Amazon S3 bucket
3. Compression seems to be possible (not default?) which is great; my backup seems not compressed though. I will check this out later.

I did not try to recover from backup which is the important part When I find the time I will do a simulated recovery. When Duplicati lives up to its promise (which I do not doubt) it is exactly what I was looking for all the time…

I would like to see it included in Xubuntu (my favorite distro) by default – that would save a lot of people a lot of headache, I suppose….

I’m a little surprised that the waist of the cph spline is so narrow (se the above example) – I can’t understand quite why this is…

I’ve updated the pastebin to the latest version.

A little off-topic – I’ve noticed some problems with the prop. hazards assumptions in my data and I want to do a time-spline the way the coxvc package did but I haven’t had any luck with my post on CrossValidated – perhaps you have an idea? I would like to have a similar plot to the plotHR but for the time axis – the plot that the timecox function provides just seems unappealing to the eye and I believe that the time function in my case is smooth…

A few days ago a Ubuntu (Debian) package has been released. http://www.duplicati.com/news/packagesforubuntudebianredhatandothers

Try it out and tell us what you think!