Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

st: Stacked Cumulative Incidence Plots

 From 千早 ケンジ To statalist@hsphsun2.harvard.edu Subject st: Stacked Cumulative Incidence Plots Date Tue, 14 Dec 2010 01:31:22 +0800

```Dear list,

I am estimating a competing risks model using stcrreg for a dependent variable that has two kinds of failure.  Using the following commands:

stset age_r, failure(homogamy==2) id( serial)
stcrreg i.cohort  i.edu_r  opportunity, compete(homogamy = 1)

Then I predict the Baseline Cumulative Incidence Function using:

predict cif0_homogamy, basecif
gen cif_cohort2_homogamy = 1 - (1-cif_homo_reforms0)^exp(_b[2.cohort])

Next I repeat these commands but stset my data for the other possible failure and predict the CIFs for that outcome.

Then I create the sum of the cause specific CIFs in order to make stacked cumulative incidence plots:

gen sum_cif_cohort1 = cif0_homogamy + cif0_heterogamy
gen sum_cif_cohort2 = cif_cohort2_homogamy + cif_cohort2_heterogamy

The problem is that when I make the stacked cumulative incidence plot I find out that the sum of the CIFs is bigger than 1.00 and the plots show probabilities of 1.65 or more. Am I wrong or these CIFs are supposed to sum to 1.00? What should I do in order to obtain stacked plots showing the probabilities of each type of failure and of no event stacked and summing up to 1.00?

Guilherme Kenji Chihaya

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```