Joinpoint Regression

[Masoumeh Sanagou]

STATALIST,

I want to fit 6 regression models with 0,1,2,3,4,5 joinpoints (knots), but the location of the joinpoints are not known and are to be estimated from the data.

“Joinpoint Trend Analysis Software” was recommended but I want to do it using STATA.


Regards,

[Leonardo Guizzetti]

I'm sure that you *could* do it in Stata, but if you're going for the same type of algorithms implemented in the Joinpoint Trend Analysis Software from the National Cancer Institute of the NIH, you'll have to do much of the programming.

That said, there are a couple of options. You could use the -loghockey- package (-search loghockey-) which is a user-contributed command that fits piecewise-linear regression models. I've not used this, and it seems like it would not fit your purpose, as it only handles one breakpoint.

The built-in command -mkspline- can be used to fit linear splines with user specified knots, which would give you the building blocks to rolling your own joinpoint regression.

In my opinion, I would not "reinvent the wheel" if I don't need to.

[Masoumeh Sanagou]

Thanks for the reply.

Is npregress series equivalent to the algorithms implementation in “Joinpoint Trend Analysis Software”?

[Leonardo Guizzetti]

I'm not familiar with -npregress- but I'm not aware of anything in Stata or from other software that are the same as what Joinpoint implements.

[Maarten Buis]

npregress is not equivalent in the sense that it is mathematically the same model, but it has a similar purpose: to fit a flexible non-linear relationship. Whether or not it is similar enough for your purpose depends on your purpose.