Harrell's C concordance statistic after streg

[Kylie Crawford]

Hi, I am wanting to calculate Harrell's C after an exponential regression model. I have opted for a flexible parametric model because the PH assumption was violated (and categorizing/transformation did not resolve and I have not used splines before). I see estat concordance is not available in the post-estimation commands after streg. I thought the Harrell's C statistic was appropriate after Weibull/Exponential- maybe I have misunderstood. I could not find a user-written package that might help. I would greatly appreciate any advice. Thanks, Kylie

[Clyde Schechter]

Following -streg- you can use -predict- to calculate a predicted mean or median survival time. Then Roger Newson's -somersd- package, available from Stata Journal, can use that and the observed survival time to calculate the Harrell C-statistic when you specify the -transf(c)- option.

[Kylie Crawford]

Great. Thanks very much. Appreciate your quick reply and help. Kylie