Creating graphs showing survival curves with confidence intervals on multiple record data

[Chris Boulis]

Hi Statalist.

Is there a way to show survival curves with confidence intervals for multiple-record data? I have been using -stcurve- for my graphing but there is not an option to display confidence intervals to survival functions. On this, the -stcurve- documentation states

See Cefalu (2011) for a Stata command to plot the survivor or cumulative hazard function with pointwise confidence intervals.

When I tried to use -survci- Stata provided this response:

. survci
survci may not be used with multiple-record data
r(498);

Help appreciated.

[Paul Dickman]

stpm2 (from SSC)

[Chris Boulis]

Thank you Paul Dickman. I read over the help file for this, while I found it complicated (I did not understand the concept of "knots" and what needed to be entered there.. I did run a basic regression and found it provided very similar hazard ratios to that when using -stcox-, but I didn't see any guidance/examples on creating the survival curves with CIs. If you know of any links with some basic examples, could you please post? Kind regards, Chris

[Paul Dickman]

Here's an example of plotting survival curves with CIs
https://pclambert.net/software/stpm2/stpm2_timevar/

This tutorial shows how to produce standardised survival curves with CIs. This is awesome!
https://pclambert.net/software/stand...ized_survival/

This covers the basics
https://pclambert.net/software/stpm2/comparewithcox/

[Chris Boulis]

I would like to produce a graph displaying survival curves showing the mean survival with 95% confidence intervals for categories of a variable. Here's an example of what I want to achieve using sts graph

sts graph, survival by(relig) ci name(relig2, replace) ///
graphregion(color(white) icolor(white)) plotregion(style(none) color(white) icolor(white)) ///
plot2opts(lcolor(maroon)) plot4opts(lcolor(orange_red)) ///  
title("Mean with 95% Confidence Intervals", size(medsmall)) ///
ytitle("Survival", size(small)) ylabel(.84(.02)1.0 , angle(horizontal)) yscale(noline) ylabel(, labsize(small)) ///
xlabel(, labsize(vsmall)) xmlabel(##5) xtitle("Duration (in Years)", size(small)) ///
legend(region(lstyle(none)) size(small) row(1) pos(6) rowgap(1) span ///
order(8 "group4" 6 "group3" 4 "group2" 2 "group1"))

which provided the following graph, but it lacks clarity with regard to the mean and their 95% CIs.

I installed stmp2 (from SSC) and after reading the help file, I ran the following

stset wave, failure(end==1) id(couple) origin(begin==1) exit(time .)
stpm2 rel, scale(hazard) df(5) eform

I have not worked out how to use -predict- to my variable with 4-categories to create the confidence intervals. I tried the following with variations

predict relsurv, at(rel(1 2 3 4)) survival ci timevar(wave)

though Stata reported errors, e.g.,"2 invalid name". Some help with that and I should then be able to graph using something like:

twoway (rarea relsurv_lci relsurv_uci wave) ///
(line relsurv wave) ///
, legend(order(1 "rel1" 2 "rel2" 3 "rel3" 4 "rel4") ring(0) cols(1) pos(1)) ///
ylabel(, angle(h) format(%3.1f)) ytitle("") xtitle("") name(ci, replace)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float couple byte(rel wave begin end)
 4 1 1 1 0
 4 1 2 0 0
 4 1 3 0 0
 4 1 4 0 0
 4 1 5 0 0
 4 1 6 0 0
 4 1 7 0 0
 4 1 8 0 1
 6 2  1 1 0
 6 2  2 0 0
 6 2  3 0 0
 6 2  4 0 0
 6 2  5 0 0
 6 2  6 0 0
 6 2  7 0 1
 7 4  1 1 0
 7 4  2 0 0
 7 4  3 0 0
 7 4  4 0 0
 7 4  5 0 0
 7 4  6 0 0
 7 4  7 0 0
 9 4  1 1 0
 9 4  2 0 0
 9 4  3 0 0
 9 4  4 0 0
 9 4  5 0 0
 9 4  6 0 0
 9 4  7 0 0
 9 4  8 0 0
 9 4  9 0 0
13 4  1 1 0
13 4  2 0 0
13 4  3 0 0
13 4  4 0 0
13 4  5 0 0
13 4  6 0 0
13 4  7 0 0
13 4  8 0 0
13 4  9 0 0
end

Stata 15.1. Using panel data.