Multiple failures in competing risk model (stcrreg)

[Rebecca Sansale]

Hi everyone,

I am working with panel data (~5 years) of employment outcomes (number of weeks in labor force, dropping out of lf, employed, unemployed) for about 9000 individuals. I have aggregated the weekly data to create a table that is unique by employment outcome spell.
So, each time an individual changes from one employment status to another, the number of weeks of that spell is totaled and then I collapse by each spell total. This multiple "failure" structure is conducive to the structure necessary to run the stcox function. While I was originally using stcox to estimate hazard ratios for these individuals, I would like to (hopefully) use a competing risks model, so that I am able to better incorporate other "risks" that young workers face while in the work force. I have read up on stcrreg, but this specific function doesn't seem to work with multiple observations per individual.

My question is-- has anyone found another competing risks model that will allow for multiple failures per individual? and if so, what were these functions?

Thank You,

Rebecca

[Ruud Gerards]

Hi everyone,

I'm struggling with the same issue, also for the analysis of labour market outcomes, and the stcreg help file explicitly states that it won't work when the data contain multiple failures per subject.

Has anyone found a solution to this or a suggestion for a workaround?

best,
ruud

[Anne Buzzell]

Hi All,

I'm wondering if perhaps someone found a way to use stcrreg with MI data? The help file on stcreg does say that it supports 'mi estimate' however, I can't seem to avoid the error stating that it cannot work due to multiple failures per subject.

Any insights would be appreciated,

Annie