what does -per(1000) - ratio in standsurv

[Denise Vella]

Hello all

I'm just practicing with standsurv.

A silly question for sure - what does per(1000) mean?
The max on the y axis is 150.... so certianly this is not % , how should I interpret the y axis?

standsurv, at1(treatment 0) at2(treatment 1) timevar(timevar) hazard ci contrast(ratio) per(1000)

Full data set from here:

clear all 
use https://pauldickman.com/data/melanoma.dta

rename surv_mm survivalt
stset survivalt, failure(status==1) exit(time 60.5) scale(12)

//create dummy variable - treament
gen treatment = 0 
replace treatment = 1 if stage == 1 
 
stpm2 treatment age stage sex, scale(hazard) df(4) eform nolog tvc(treatment sex) dftvc(3)

range timevar 0 10 100 

standsurv, at1(treatment 0) at2(treatment 1) timevar(timevar) hazard ci contrast(ratio) per(1000) //don't know what this means

twoway (rarea _at1_lci _at1_uci timevar, color(red%30)) ///
(rarea _at2_lci _at2_uci timevar, color(blue%30)) ///
(line _at1 timevar, color(red)) ///
(line _at2 timevar, color(blue)) ///
, legend(off) ///
ylabel(,angle(h) format(%3.1f)) ///
xtitle("Years from surgery")  

//Should be similar to: 
sts graph, hazard by(treatment) kernel(epan) name(h,replace)

[Paul Dickman]

Hazard is a rates; events per 1 time unit. per(1000) rescales to events per 1000 time units.

in this example it’s deaths per 1000 person-years.

surv_mm is time in months. scale(12) in stset rescales to make the time unit 1 year.