Drawing generic survivorship curves

[Andrew Lover]

Hi all,

I am working to draw some illustrative example survivorship curves as in the diagram attached. Note, these are slightly different form 'normal' survival curves in that everything goes to zero.

* simplest option?

// clear all

set obs 1000

gen t = _n
gen fail =  rbinomial(1, 0.1)
tab fail

stset t, fail(fail)

sts graph, title("")
sts gen surv1 = s
twoway (lowess surv1 t, sort lpattern(solid) lwidth(thick) lcolor(orange))

* more complex options
// ssc install survsim

clear all
set obs 1000
set seed 123
gen treat = rbinomial(1,0.5)

* lambda = scale; gamma = shape
 
survsim time1 fail, lambdas (20) gammas(10) cov(treat 10) maxt(2)
stset time1, fail(fail = 1)
sts graph, by(treat) title("")

clear all
set obs 1000
set seed 123
gen treat = rbinomial(1,0.5)
survsim time1 fail, lambdas (20) gammas(0.5) cov(treat 10) maxt(2)

stset time1, fail(fail = 1)
sts graph, by(treat) title("")
exit

However, I am struggling to get smooth nice-looking curves. Any ideas on better ways to attack this problem?

thanks!

[Svend Juul]

It is not very clear what you want. But with a constant incidence rate (IR; here 0.02 per time unit T), the corresponding survival function is exp(-IR*T):

twoway function y=exp(-.02*x) , range(0 100)

Is that what you are thinking of?

[Andrew Lover]

Hello Svend,

Thanks, that's certainly much closer to what I was hunting for:

tw (function y=exp(-.065*x) , range(0 100))
tw (function y=-exp(0.065*x) , range(0 100))

It always surprises me how limited the manuals are regarding -function-; I wasn't aware you could even call it directly inside -twoway-!

cheers

[Nick Cox]

The last of three examples in the manual entry shows a combination of twoway function and scatter with data.

A tiny trick I often use for adding lines of equality over some range is

... || function equality = x, ra(xwhatever)

The text equality is then echoed to the legend.

[Andrew Lover]

Thanks Nick, that's very helpful.
Just to save others from the same confusion I had: Nick is referencing the manual entry for -twoway function- which has no links from the entry for -function- (as far as I can tell).