growthTrendR (version 0.2.2)

ci_resp: Compute prediction and confidence intervals of smooth terms on response scale

Description

This function computes predicted values and confidence intervals from a fitted GAM model with a log-link (or other link) and optionally back-transforms predictions to the response scale. Five methods are supported: 1. **delta_link**: classic delta method on the linear predictor (link) scale; back-transformed CI is asymmetric. 2. **delta_resp**: delta method applied directly on the response scale using \(\mathrm{Var}[\exp(\eta)] \approx \exp(2\eta)\,\mathrm{Var}(\eta)\). 3. **bootstrap_link**: parametric bootstrap on the linear predictor; quantiles back-transformed. 4. **bootstrap_resp**: parametric bootstrap on the response scale; quantiles computed after exponentiating. 5. **posterior**: Bayesian posterior simulation using the model covariance matrix; quantiles on response scale.

Usage

ci_resp(
  model,
  newdata,
  nboot = 100,
  method = c("posterior", "delta_link", "delta_resp", "bootstrap_link", "bootstrap_resp"),
  level = 0.95
)

Value

A data.table with columns: fit, lwr, upr.

Arguments

model

A fitted GAM object from mgcv::gam.

newdata

A data.frame containing values at which to predict.

nboot

Integer. Number of bootstrap or posterior samples. Default 1000.

method

Character. One of "delta_link", "delta_resp", "bootstrap_link", "bootstrap_resp", "posterior".

level

Numeric. Confidence level, default 0.95.

Details

References: - Wood, S.N. (2017) *Generalized Additive Models: An Introduction with R, 2nd Edition*. CRC Press. - Efron, B., & Tibshirani, R. (1993) *An Introduction to the Bootstrap*. Chapman & Hall. - Gelman, A., et al. (2013) *Bayesian Data Analysis, 3rd Edition*. CRC Press.

Examples

# loading processed data
dt.samples_trt <- readRDS(system.file("extdata", "dt.samples_trt.rds", package = "growthTrendR"))
# climate
dt.clim <- data.table::fread(system.file("extdata", "dt.clim.csv", package = "growthTrendR"))
# pre-data for model
dt.samples_clim <- prepare_samples_clim(dt.samples_trt, dt.clim)
dt.m <- dt.samples_clim[ageC >1]
# using gamm_spatial model as an example
m.sp <-gamm_spatial(data = dt.m, resp_scale = "resp_log",
       m.candidates = "bai_cm2 ~ log(ba_cm2_t_1) + s(ageC) + s(FFD)")

dt.m[, uid_site.fac:= as.factor(as.character(uid_site))]
dt.ci <- ci_resp(m.sp$model$gam, newdata = dt.m)