0

I want to prove different regression coefficient for SBP on BMI according to BMI categories.

So I want to use piecewise regression.

BMI is categorized by WHO guideline which means medicaly proven breakpoint. (<18.5 , 18.5-23, 23-25, 25-30, >30)

Can I use these breakpoint ?

Or do I have to find breakpoint statistically? Like by using segment package in R.

Koh Heebyung
  • 1
  • 1
    What do you need the brakepoints for? – Tim Sep 30 '19 at 17:45
  • What is the evidence these breakpoints are in any sense "medically proven"? BMI is continuous, it would be astonishing if there was some other continuous thing (like SBP) that would have a discontinuous relationship with BMI (or even one with several steep jumps in it). That they're medically conventional - even diagnostically useful - I have no beef with, but the claim seems considerably too strong. If you want to prove that the relationship between SBP and BMI is not constant across the whole range of BMI by showing that it differs when you use those pre-determined breakpoints, ... ctd – Glen_b Oct 01 '19 at 08:22
  • ctd... then do that, but that doesn't mean this is where the breakpoints actually are, or that there are even any real breakpoints at all; if the relationship was nonlinear but completely smooth you'd likely still reject the null with a large enough sample. Similarly if you fit a suitable unspecifed-breakpoint model (and use tests or CIs or whatever suitable for such a situation), you'll still find breakpoints with a large enough sample size (if the null is a single linear relationship), even if there are no actual breakpoints, but just a smoothly changing curve. – Glen_b Oct 01 '19 at 08:24

0 Answers0