3 Comments
User's avatar
Joshua Born's avatar

In addition to plotting residuals, many models are amenable to nonparametric or pseudo-nonparametric fits. For instance, in this example you could fit a model with a binary (1 or 0) categorical variable for each age, and then plot parameter estimates for these terms. This allows a free shape for the fit on age. This way you can get a shape for age after one fitting, rather than have to iterate over multiple fittings.

Cremieux's avatar

Trying this now.

barnabus's avatar

C, unfortunately, biologically and mathematically smart biostatisticians are hard to find. It is a bit like the proverbial Eshet Chayil: "An accomplished woman who can find her? Her value is far beyond pearls. Her husband's heart trusts in her, and he lacks nothing..."

If I instead of RFK jr were running the Health Department overseeing CDC and NIH, I would make sure that at least 5% of the grant money were going to biostatistics and nowhere else, and make a complementary demand on the university overheads. Judging from my experience at a premier research university getting good quality biostatistics expertise is not easy.