Hello R users, I have a quick question I was hoping to get your input on. I am new to R and the smooth statistical regression world, and am trying to wrap my mind around the issues concerning using splines for mixed effect modeling. My question is the following: in the ?gamm? function, generalized additive mixed models can be estimated by including random components. These can be explicitly defined in the syntax, where you can also define whether the random component is an intercept, slope, or both. My understanding is that in the gam/bam function the same is achieved by including the bs="re? option for random intercepts and linear random slopes. Am I correct? If so, is there a way to specify whether it is the intercept or slope we are interested in, and does that have any effect on the output of the model? I hope these questions make sense, and I look forward to learning more about this. Thanks for taking the time to read through this email.
Random effects in package mgcv
2 messages · Dean Force, Simon Wood
If you want to include random intercepts in a model fit by bam/gam, then include s(g,bs="re") in the model formula, where g is a factor variable with one level for each group requiring a random intercept. Now suppose that for each level of group g you want a random slope w.r.t. x. You should include a term s(g,x,bs="re") in the gam formula (the order of g and x is not important). For random slopes and random intercepts include s(g,bs="re") + s(g,x,bs="re") gam/bam only supports quite simple i.i.d random effects, so correlated intercepts and slopes can not be estimated this way. Also the methods are not efficient for thousands of random effects (actually bam(,discrete=TRUE) in recent versions will manage thousands, but not 10s of thousands). best, Simon
On 27/04/16 15:12, Dean Force wrote:
Hello R users, I have a quick question I was hoping to get your input on. I am new to R and the smooth statistical regression world, and am trying to wrap my mind around the issues concerning using splines for mixed effect modeling. My question is the following: in the ?gamm? function, generalized additive mixed models can be estimated by including random components. These can be explicitly defined in the syntax, where you can also define whether the random component is an intercept, slope, or both. My understanding is that in the gam/bam function the same is achieved by including the bs="re? option for random intercepts and linear random slopes. Am I correct? If so, is there a way to specify whether it is the intercept or slope we are interested in, and does that have any effect on the output of the model? I hope these questions make sense, and I look forward to learning more about this. Thanks for taking the time to read through this email. [[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Simon Wood, School of Mathematics, University of Bristol BS8 1TW UK +44 (0)117 33 18273 http://www.maths.bris.ac.uk/~sw15190