Interpreting Stset results

[Hayoung Choi]

Hello, I'm having a problem interpreting the stset results.

I have multiple-record data, that observations were collected over 11 years to 262 persons (the survey started with 262 subjects in the first year)
and down to 158 subjects in the last year.
The dataset are like below:

year	Freq.	Percent	Cum.
2009	262	11.58	11.58
2010	238	10.52	22.10
2011	234	10.34	32.45
2012	221	9.77	42.22
2013	211	9.33	51.55
2014	205	9.06	60.61
2015	195	8.62	69.23
2016	188	8.31	77.54
2017	180	7.96	85.50
2018	170	7.52	93.02
2019	158	6.98	100.00
Total	2,262	100.00

The failure happens when the subeject exits from medical welfare benefit (exit=1 / remain=0)
and all subjects started as medical welfare recipients in the dataset (medical_final=0).
When I do this: stset tf, failure(medical_final=1) id(h_pid).
it says that "390 observations begin on or after (first) failure.
Does it mean that it will exlude those 390 observations for the survival analysis?
and if so, how do i fix the command if I want to include those 390 observations.
Should I assign time frame?

Please, refer below for more detail:

stset tf, failure(medical_final=1) id(h_pid)
id: h_pid
failure event: medical_final == 1
obs. time interval: (tf[_n-1], tf]
exit on or before: failure
2,262 total observations
390 observations begin on or after (first) failure
1,872 observations remaining, representing
262 subjects
87 failures in single-failure-per-subject data
1,891 total analysis time at risk and under observation
at risk from t	=	0
earliest observed entry t	=	0
last observed exit t	=	11

[Paul Dickman]

Does it mean that it will exlude those 390 observations for the survival analysis?

Yes, those observations will be excluded.

how do i fix the command if I want to include those 390 observations

This depends on what the issue is. If these 390 observations have all failed within the first year, and have been assigned a survival time of 0 completed years, then the solution is to add a small amount to their survival times. If the issue is something else (e.g., observations with negative survival time for some other reason) then you need another solution.

See the following blog by Bill Gould

https://www.stata.com/support/faqs/s...and-cox-model/

When it comes to analysis, you may want to look at methods for discrete time data.

[Hayoung Choi]

Yes, those observations will be excluded.



This depends on what the issue is. If these 390 observations have all failed within the first year, and have been assigned a survival time of 0 completed years, then the solution is to add a small amount to their survival times. If the issue is something else (e.g., observations with negative survival time for some other reason) then you need another solution.

See the following blog by Bill Gould

https://www.stata.com/support/faqs/s...and-cox-model/

When it comes to analysis, you may want to look at methods for discrete time data.

thank you so much for your response! I will replace time for those who failed within 0 time.
thank you again.