stlmm: Fit a Spatio-Temporal Linear Mixed Model
In michaeldumelle/DumelleEtAl2021STLMM: Fit Spatio-Temporal Linear Mixed Models

Fit a Spatio-Temporal Linear Mixed Model

stlmm(data, formula, ...)

## S3 method for class 'data.frame'
stlmm(
  data,
  formula,
  xcoord,
  ycoord = NULL,
  tcoord,
  stcov,
  estmethod = "reml",
  s_cor = "exponential",
  t_cor = "exponential",
  chol = FALSE,
  condition = 1e-04,
  logdet = FALSE,
  weights = "cressie",
  initial = NULL,
  optim_options = NULL,
  h_options = NULL,
  max_options = NULL,
  stempsv_options = NULL,
  ...
)

`data`	A data object containing all necessary variables.
`formula`	A formula of the form `y ~ x`, where `y` is the response variable and `x` are the predictor variables.
`...`	Additonal arguments.
`xcoord`	A character vector specifying the column name of the x-coordinate variable in `data`.
`ycoord`	A character vector specifying the column name of the y-coordinate variable in `data`.
`tcoord`	A character vector specifying the column name of the t-coordinate (time) variable in `data`.
`stcov`	The spatio-temporal covariance type `product` The product LMM `sum_with_error` The sum-with-error LMM `productsum` The product sum LMM
`estmethod`	The estimation method `reml` Restricted Maximum Likelihood `svwls` Semivariogram Weighted Least Squares
`s_cor`	The spatial correlation `exponential` The exponential correlation (the default). `spherical` The spherical correlation. `gaussian` The Gaussian correlation.
`t_cor`	The temporal correlation `exponential` The exponential correlation (the default). `spherical` The spherical correlation. `gaussian` The Gaussian correlation. `tent` The tent (linear with sill) correlation.
`chol`	Should the Cholesky decomposition be used? If `FALSE`, efficient inversion algorithms are implemented. Defaults to `FALSE`.
`condition`	A small number added to the diagonals of matrices before inverting them to prevent ill-conditioning (defaults to `1e-4`).
`logdet`	Should the log determinant be returned? (defaults to `FALSE`).
`weights`	Weights when `estmethod = "svwls"` (defaults to `"cressie"`) for the Cressie's weighted least squares weights).
`initial`	Initial values for the parameters. Must be made with `make_covparam_object()` (defaults to even spread of across variance parameters, a total variance matching the sample variance of OLS residuals, and ranges equaling half the maximum distance in the domain).
`optim_options`	A list containing additional options to pass to optim.
`h_options`	A list containing options to compute distances if `response`, `xcoord`, `ycoord`, and `tcoord` are provided. Named arguments are `h_t_distmetric` The temporal distance matrix (defaults to `"euclidean"`). `h_s_distmetric` The spatial distance matrix (defaults to `"euclidean"`).
`max_options`	A list containing additonal options for placing upper bounds on the total variance, spatial range, and temporal range. This can be helpful for numerical stability in optimization. `max_v` The maximum total variance (defaults to four times the variance of OLS residuals) `max_s_range` The maximum spatial range variance (defaults to four times the maximum observed spatial distance) `max_t_range` The maximum temporal range variance (defaults to four times the maximum observed temporal distance)
`stempsv_options`	A list containing additional options for the empirical spatio-temporal semivariogram. Named arguments are `n_s_lag` The number of spatial distance classes (defaults to 16). `n_t_lag` The number of temporal distance classes (defaults to 16). `h_s_max` The maximum spatial distance. Deafaults to half the maximum distance in the spatial domain. `h_t_max` The maximum temporal distance. Deafaults to half the maximum distance in the temporal domain.

A list containing several objects

CovarianceParameters: Estimated covariance parameters.
Coefficients: Fixed effect estimates.
NamesCoefficients: Names of the fixed effect estimates.
CovCoefficients: The covariance matrix of the fixed effect estimates.
Objective: A list containing optimization information.
CovarianceForms: The spatial, temopral, and spatio-temporal correlation forms.
formula: The model formula.
model: A list containing the fixed effect design matrix and response vector.
data_object: An ordered data object.
invert_object: An inverse object.
coord_names: The names of the coordinate vectors.
coords: The coordinate vectors.
h_options: Returning the h_options argument.
stempsv_options: Returning the stempsv_options argument.
stempsv: The empirical spatio-temporal semivariogram (if estmethod = "svwls")
optim_options: Returning the stempsv_options argument.
max_options: Returning the max_options argument.
chol: Returning the chol argument.
condition: Returning the condition argument.
residuals: Raw residuals.