Survival analysis with spell-type data

[Will Fannin]

I am trying to do survival analysis with program participation data. Here is my data:

* Example generated by -dataex-. For more info, type help dataex
clear
input long uniqueid float modate double program byte(_seq _spell _end) int(_start _enddate)
1 696 1  7 1 0 690 708
1 697 1  8 1 0 690 708
1 698 1  9 1 0 690 708
1 699 1 10 1 0 690 708
1 700 1 11 1 0 690 708
1 701 1 12 1 0 690 708
1 702 1 13 1 0 690 708
1 703 1 14 1 0 690 708
1 704 1 15 1 0 690 708
1 705 1 16 1 0 690 708
1 706 1 17 1 0 690 708
1 707 1 18 1 0 690 708
1 708 1 19 1 1 690 708
1 742 1  1 2 0 742 743
1 743 1  2 2 1 742 743
2 696 1 25 1 0 672 702
2 697 1 26 1 0 672 702
2 698 1 27 1 0 672 702
2 699 1 28 1 0 672 702
2 700 1 29 1 0 672 702
2 701 1 30 1 0 672 702
2 702 1 31 1 1 672 702
2 708 1  1 2 0 708 742
2 709 1  2 2 0 708 742
2 710 1  3 2 0 708 742
2 711 1  4 2 0 708 742
2 712 1  5 2 0 708 742
2 713 1  6 2 0 708 742
2 714 1  7 2 0 708 742
2 715 1  8 2 0 708 742
2 716 1  9 2 0 708 742
2 717 1 10 2 0 708 742
2 718 1 11 2 0 708 742
2 719 1 12 2 0 708 742
2 720 1 13 2 0 708 742
2 721 1 14 2 0 708 742
2 722 1 15 2 0 708 742
2 723 1 16 2 0 708 742
2 724 1 17 2 0 708 742
2 725 1 18 2 0 708 742
2 726 1 19 2 0 708 742
2 727 1 20 2 0 708 742
2 728 1 21 2 0 708 742
2 729 1 22 2 0 708 742
2 730 1 23 2 0 708 742
2 731 1 24 2 0 708 742
2 732 1 25 2 0 708 742
2 733 1 26 2 0 708 742
2 734 1 27 2 0 708 742
2 735 1 28 2 0 708 742
2 736 1 29 2 0 708 742
2 737 1 30 2 0 708 742
2 738 1 31 2 0 708 742
2 739 1 32 2 0 708 742
2 740 1 33 2 0 708 742
2 741 1 34 2 0 708 742
2 742 1 35 2 1 708 742
end
format %tm modate
format %tm _start
format %tm _enddate

In order, the variables are a person identifier, month-date, indicator for program participation, tspell output variables (_seq, _spell, _end), and start and end dates for each spell.

My initial approach was the following:

stset _seq, id(uniqueid) failure(_end) exit(time .)

But the FAQ for survival analysis with spell-type data (https://www.stata.com/support/faqs/statistics/stset-spell-type-data/) seems to suggest something like the following:

stset _enddate, failure(_end) origin(_start) id(uniqueid) exit(time .)

Output from neither approach looks correct.