Hello R-INLA Community,
Getachew Dejene
Hello R-INLA Community,
I’m working with a "spatio-temporal" dataset structured as follows:
My goal is to model seasonal patterns with a 131-week period across 21 counties, and I’m incorporating trigonometric functions as a function of time (using week to capture the seasonal variation). The model I’m aiming to fit is as follows:
Where:
- : Baseline effect for each county (random intercept).
- and : Coefficients for sine and cosine terms to capture seasonality.
- : Spatial effect for each county.
- : Temporal random effect for each week.
- : Residual error term.
To achieve this, I specified the model formula in R-INLA as:
#######
formula_trig <- cases ~ sin_week + cos_week + f(County, model = "iid") +
f(ID.County, model = "bym", graph = Al_g) + f(week, model = "rw1")
########
model <- inla( formula_trig, family = "poisson",
data = df, control.predictor = list(compute = TRUE), control.compute = list(dic = TRUE, waic = TRUE)
, Since I’m relatively new to R-INLA, I’m reaching out for guidance on:
- Best practices for using trigonometric functions within a spatio-temporal model in INLA.
- Model setup advice, especially around specifying spatial and temporal random effects.
- Any recommended resources or tutorials that could help me understand how to incorporate time-based trigonometric functions for seasonal patterns.
Any insights, examples, or references to relevant materials would be greatly appreciated!
Thank you very much for your time and help!
Best regards!!