[R-sig-ME] glmmTMB- fitting splines
John Maindonald
john@m@indon@ld @ending from @nu@edu@@u
Tue May 22 07:00:21 CEST 2018
The spline coefficients multiply the two basis terms. Most times, one wants to
work with predicted values and standard errors. The predict method seems
not yet to have been implemented for glmmTMB models with a betabinomial
error family. One can use fitted() to get just the fitted probabilities, and do a
complementary log-log transform (in this instance) to get predictions on the
scale of the linear predictor (NB, linear in the sense that it is a linear combination
of the basis functions, plus intercept).
The output suggests that the first basis term might be enough on its own.
Observe, however, to choose a simple case:
> x <- 1:5; splines::ns(x, 2)[, 1]
[1] 0.0000000 0.3570466 0.5662628 0.5290951 0.3440969
This is a very nonlinear function of x, quite different from the linear function
of x that one gets by typing splines::ns(x, 1)
Regression thin plate splines, as implemented in mgcv. have the advantage
that the initial basis terms change only very slightly as one moves to a higher
degree of freedom basis.
John Maindonald email: john.maindonald at anu.edu.au<mailto:john.maindonald at anu.edu.au>
On 22/05/2018, at 14:41, dani <orchidn at live.com<mailto:orchidn at live.com>> wrote:
Hi John,
Thank you so much! This is very helpful! I managed to run it but I am not sure how to interpret the results as I get this:
# Conditional model:
# Estimate Std. Error z value Pr(>|z|)
# (Intercept) -8.40461 1.58077 -5.317 1.06e-07 ***
# splines::ns(newage, 2)1 -1.89262 0.57246 -3.306 0.000946 ***
# splines::ns(newage, 2)2 0.10296 0.47268 0.218 0.827575
I am not sure what to make of the two different spline results.
Best regards,
D
________________________________
From: John Maindonald <john.maindonald at anu.edu.au<mailto:john.maindonald at anu.edu.au>>
Sent: Monday, May 21, 2018 7:10 PM
To: dani
Cc: r-sig-mixed-models at r-project.org<mailto:r-sig-mixed-models at r-project.org>
Subject: Re: [R-sig-ME] glmmTMB- fitting splines
There is an example at http://www.rpubs.com/johnhm/Overdispersed
See Section 2.2 .
John Maindonald email: john.maindonald at anu.edu.au<mailto:john.maindonald at anu.edu.au>
On 22/05/2018, at 11:36, dani <orchidn at live.com<mailto:orchidn at live.com>> wrote:
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list