Cumulative incidence function for competing risks discrete time analysis

[Chris Rooney]

So I've searched for about 3 hours on the internet, using all sorts of search terms, but it has borne no fruit.

I have a competing risks discrete time analysis model (coded 1, 2, 3, 4 for the various events, and "0" for censored observations). As a result, I cannot use the Kaplan-Meier survival curve and as an alternative, the Cumulative Incidence Function (CIF) was suggested. However, all the stata sources that I look up assume the data are continuous. I cannot find a single example (even on Stephen Jenkins' site) of a CIF with discrete time data. Does anyone know how to do this and could you please provide the steps to do so (any generic example will do)?

[Stephen Jenkins]

Thanks for the citation. However, your question is a sore point with me, as your question is something I've been meaning to look at for a long time and not got around to! My first reaction is: can't one go back to "first principles" and use the discrete time analogue of the relevant continuous time formula? For instance, have a look at the formulae on pages 2398-2399 of H. Putter, M. Fiocco1 and R. B. Geskus, "Tutorial in biostatistics: Competing risks and multi-state models", Statistics in Medicine, 2007; 26:2389–2430. In short I think you need estimates (for each interval) of the overall survivor function, and of the cause-specific hazard of interest -- and these are derivable, but you'd need to assemble them. Confirmation from experts would be appreciated! And also literature references.