Generalized Additive Models

To fit a GAM in R using mgcv, specify your model formula with smooth terms like s(time) and s(temperature), and set the family argument to quasipoisson. This models non-linear relationships effectively within your dataframe.

Concurvity in GAM diagnostics measures the degree of collinearity between smooth terms, similar to multicollinearity in linear models. You check concurvity using mgcv's built-in diagnostic tools to identify overlapping smooth term patterns.

Yes, mgcv supports both REML and GCV for smoothing parameter selection in Generalized Additive Models. You specify the method argument in the gam function, with REML generally recommended for more stable penalization performance.

Assess basis dimension adequacy by checking the effective degrees of freedom and diagnostic outputs in mgcv. If the effective degrees of freedom approaches the maximum basis dimension, you should increase the k parameter for that smooth term.

Yes, mgcv integrates with DLNM to fit penalized distributed lag models. This allows you to model exposure-lag-response relationships while applying automatic penalization to the smooth lag terms.

Use Generalized Additive Models when your data exhibits non-linear relationships that standard linear models cannot capture. GAMs fit flexible smooth functions to predictors, allowing you to model complex curves without manually specifying polynomial degrees.