Dear all,

Does anyone know how to crop (or not show in some way) the highest part of the yaxis in a sts graph? Let me explain my data and problem:
I am using sts graph, hazard to estimate risk of a type of cancer (cancer==1) in participants who are exposed (exposed==1) or unexposed (exposed==0) to a certain risk factor. I have a large dataset with about 3,330,000 participants, and I since the exposure is time-dependent, I have split the file (using stsplit), so there are 6,622,000 observations in the dataset.
As the data is sensitive, I have to use Stata on a secure server where I cannot copy anything or use the internet, so I am unfortunately not able to use dataex. In addition, I wanted to attach the graphs I have made, but I am not allowed to export graphic files (only pdfs) and when I took a picture of the graph with my phone, saved it as a .png-file and tried to upload it, I got ; "Error uploading image"). I hope someone is still able to help me.

I want to compare two graphs, one using stset with birthyear as origin, and one using study entry as origin.
stset date_eof_cancer, failure(cancer==1) entry(study_entry) origin(birth_year) sc(365.25) id(id)
and
stset date_eof_cancer, failure(cancer==1) entry(study_entry) origin(study_entry) sc(365.25) id(id)

My code for the graph is:

sts graph, by(exposed) hazard ci ///
ylabel(0 "0" 0.0001 "10" 0.0002 "20" 0.0003 "30" 0.0004 "40" 0.0005 "50") /// //and in addition 0.001 "100" 0.0015 "150" for the one with birth_year as origin
xlabel(0(5)35) // and some other options for esthetics which I suppose is not relevant to the question

The problem is that when I use the birth_year as origin, the confidence interval in the start of the curve for the exposed group is very high (150 cases per 100 000 person-years) as compared to the rest of the two graphs, where the maximum values at the yaxis is about 50 cases per 100 000 person-years.
As I would like to compare the two graphs and use them in a publication, I would prefer if I could have the same scale and labels on the yaxis. The relevant information would still be included in the graphs, and I would write a foot-note to the table describing the maximum value of the confidence interval (unless there is a way to show it in the graph without showing the whole curve of the confidence interval). As you probably know, it does not help to change the ylabel.

I have searched the Statlist forum and other websites, but not found the right way to do this. Is it even possible?
In one previous post from the Statalist (https://www.stata.com/statalist/arch.../msg00224.html), David Harrison writes "If you want to restrict the range so that some of the curve is not visible, you will need to use -sts gen- to generate the full curve and then plot only the bit you want using a standard twoway graph." Maybe that is what I need to do? However, I have not managed to make the graph this way (I can give more information about how I have tried doing this, if this is the best way to solve my problem).

I am thankful for any help,

Best regards,
Marie S. Sandvei
Resident physician and researcher