08-28-2013 10:14 AM
I gather there is no such ODS Table. So does anyone know how the Confidence Intervals are calculated for the Quartiles when expressed in time? We are trying to replicated the ODS Table for Quartiles and expand it for additional timepoints (10%, 20%, etc.).
08-28-2013 10:36 AM
Tables that might get what you want are ODS LifetableEstimates/ProductLimitEstimates, or the OUTSURV= dataset depending on the method used. I think those estimates should have the CI around them, if not the definition is in the manual, under details>computation formulas, but it depends on the method specified.
The data in the quartiles table will correspond to one of those tables as well, though the exact one escapes me at the moment.
08-28-2013 11:23 AM
Could you provide a sample data and program to tinker with?
In theory, you can modify the template for stat.lifetest.quartiles table to achieve the 10%, 20% etc setup directly with the existing template rather than rebuilding from scratch using the same method to calculate your intervals manually.
I'm not yet familiar enough with proc template and I am still waiting for the proc template book to arrive back at the library at work so I can't be of much further help but if you can provide a working small set of data (it can be one of sashelp datasets) with a basic proc lifetest, I would gladly try to tinker around and provide you with findings.
*at least that's what I understood from proc template
08-28-2013 12:01 PM
I don't think you can just change the template here as if there are values that are missing, eg. the curve never actually hits 25% it uses the next largest value, so the data has to be prepared first I would guess.
EDIT: You'd be changing the source data is what I mean to say, and I suppose its possible, but seems way more work than necessary.