Summarise a gamRTMB fit

Laid out like mgcv::summary.gam(), but with one row per distributional parameter throughout, since every parameter has its own formula, link, coefficients and smooths.

Usage

# S3 method for class 'gamRTMB'
summary(object, ...)

# S3 method for class 'summary.gamRTMB'
print(x, ...)

Arguments

object

A gamRTMB fit.

...

Ignored.

x

A summary.gamRTMB.

Value

An object of class summary.gamRTMB.

Details

Parametric coefficients get standard errors from vcov.gamRTMB() and approximate Gaussian (z) tests. Smooth terms are reported with their effective degrees of freedom and smoothing parameters but no p-values: mgcv's "approximate significance" rests on Wood's (2013) test for a term's whole coefficient block, which is not implemented here, and a naive Wald test in its place would be misleading.

Examples

set.seed(1)
d <- data.frame(x1 = runif(300), x2 = runif(300))
d$y <- rnorm(300, sin(2 * pi * d$x1), exp(-1 + d$x2))
summary(gamRTMB(y ~ list(mean = ~ s(x1), sd = ~ s(x2)), data = d))
#> 
#> Family: norm   [dnorm from RTMB]
#> Links:  mean = identity,  sd = log
#> 
#> Formula:
#>   mean ~ s(x1)
#>     sd ~ s(x2)
#> 
#> Parametric coefficients:
#>                  Estimate Std. Error z value Pr(>|z|)    
#> mean:(Intercept)  0.05266    0.03374   1.561    0.119    
#> sd:(Intercept)   -0.45491    0.04092 -11.118   <2e-16 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
#> Smooth terms:
#>        term   edf k        sp
#>  mean:s(x1) 6.377 9    0.1225
#>  sd:s(x2)   1.000 9 2.828e+08
#> 
#> Total EDF = 9.38   n = 300
#> -REML = 309.343   logLik = -289.210   AIC = 597.17