3

I'm trying to study fluctuations of a disease's activity over time, for example the f;uctiation in severity of chronic pain (in the absence of obvious triggers).

Individuals generally demonstrate one of a number of trajectories; eg. those with constant severe pain, those with minimal pain, those with flares once/year, those with several flares, pattern-less individuals etc.

I've noticed several existing studies grouping them subjectively, by appearance (see figure).

enter image description here

Is there an objective / statistical approach to group and/or describe these patterns? (for a novice)

Thank you

bobmcpop
  • 1,063
  • 1
  • 14
  • 20
  • So a statistician referred me to a text by Axen et al 2012 which recommended the following options 1) visual (as in above figure) with poor inter-observer reliability; 2) cluster analysis using hierarchical methods; 3) latent class analysis. And many others were mentioned that I don't recognize: artificial neural networks, probabilistic data mining. Anyone care to comment? – bobmcpop Nov 19 '16 at 12:10

1 Answers1

0

For those who are still looking for an answer, there are several answers on CV that address similar questions:

https://stats.stackexchange.com/a/2781/112640

How to find groupings (trajectories) among longitudinal data?

Community
  • 1
bobmcpop
  • 1,063
  • 1
  • 14
  • 20