autoplot | Automatic plotting with ggplot | Data Visualization library

This does seem to be a bug in ggfortify. As a temporary fix, you can set the survival percentage at t = 0 to 100% by doing:

autoplot comes from ggplot2. So you will need to use the ggplot2 functions to adjust items from the autoplot.

E.g. for adjusting the title size:

One option is to extract the ggplot figure from a ggsurvplot object.

In the autoplot method, the principal components are scaled, so to get the same result you would do:

The problem with using autoplot is that, although it is great for producing nice visualizations of common data structures and models with little effort, it doesn't give you the full freedom to customize the plot. However, it is pretty straightforward to do the whole thing within ggplot. The following is a full reprex:

If you specify inherit.aes = FALSE in geom_ribbon() you avoid that specific error, i.e.

Thanks for the example, I was able to get it to run without any errors, as follows:

if by continuous you mean that the black line is not "connected" to the red nor the blue: each line has a first and last observation which are connected via lines (and have possibly more observations).

Your first value to predict (or test) is, chronologically speaking, after the training data, thus the black line will not reach the blue and red. Also the blue and red line can not reach back to connect as the last observation from training data is not included in theire data.

After forecasting you could include the last observation from the training data in test and prediction set. This will lead to a "continuous" line when plotting.

There are a few things to point out here. Firstly, autoplot is a generic function, so the method used and the arguments it accepts depends on the type of object you are passing to it. In this case, you are passing a survfit object, and you will be able to see the correct parameters to use if you type ?autoplot.survfit into the console.

From this you will see that there is no legendLabs or plotTable option, and that the alpha for the confidence intervals is controlled with conf.int.alpha = . Similarly, the censoring shape is controlled with censor.shape.

Another issue is that there doesn't seem to be a way to change the factor labels in the legend, but in this case it is easy enough to change them in the data when you create the survfit object.

Lastly, it's a good idea to make a reproducible example if you want prompt and useful answers. It took a while to recreate a reasonable data structure to test and demonstrate this answer.