Character scalar. Specifies the spline type which should be fitted. This can be either thin plate splines (tp.spline), penalised B-splines (i.e, P-splines p.spline) or smoothing splines (smooth.spline).

Integer scalar. Determines the number of knots. Per default, the number of knots is chosen adaptively to the number of unique observations. The default number also depends on the spline type.

Integer scalar. Specifies a constant multiplier to inflate the degrees of freedom in the "GCV" method to increase penalisation of models that are too close to the data and thus not smooth enough.

Character scalar. The smoothing parameter estimation method. Currently, only "REML", code"ML" and "GCV" are supported. This argument is ignored for type = "smooth.spline". For details see gam and gamm (see package mgcv).

list. Further arguments to be passed to the smoother s (see package mgcv). Note that the mgcv options k and bs are specified using type and knots in opm.

An optional "corStruct" object (see the help topic corClasses in the nlme package) as used to define correlation structures in package nlme. For better coverage of confidence intervals and slightly improved spline fits it is advised to use an AR process of order 1 or 2. However, this correction for auto-correlated error terms results in increased run time.

Should the models be saved (on the disk) for further inspections and plotting?

File name of the models. Per default a name is auto-generated based on date and time. The file is always generated in the current working directory.

Additional arguments to be passed to gam or smooth.spline.