[notes] Meeting notes (R code for Mixed Effects models)

December 03, 2014

Notes from today's meeting with Emily

Modelling repeated measure data for diads (e.g coupled osc)

standard: multi-level model papers that introduced it?

  most doable in sass and R, 
      one piece in sass but not R

Weighted sum of predictors multi-level - fixed effects model global average plus individual average

What happens to residuals around lines.

eg. my result is grand slope plus my difference, plus residual

"what happens to residuals?" what does that mean model residuals auto-correlation or partner covariation want both hack in R: put them in as fixed effects. lag-1 <- don't totally understand this R package - NLME nonlinear mixed effects models -> lme() function is most used -> correlation structure -> Type A: "variance functions" -> NLME defines a default set of variance functions -> Type B: "correlation structures" -> NLME defines a default set of CorrStruct classes We may need to define our own SASS version - Emily will send the url - a "repeated statemnt" sets up structure on residuals (the "R matrix") - "type equals" the type we want in SASS-speak un@ar1 "direct product AR 1"

S or S+ are "paid versions" of R.


Where can I get sass? runs in VMWare runs through the internet through UA Emily has it running


example data example results in SASS

book: mixed effects models in S or S+ . pinhiero bates niell: longitudinal data in R

Posted by Kyle Simek
blog comments powered by Disqus