Given a fitted model and additional information, this function computes the corresponding internally studentized residuals.
rstudents(model, varobj, method = c("Richardson", "simple", "complex"), ...)
# S4 method for missingORNULL
rstudents(
model,
varobj,
method = c("Richardson", "simple", "complex"),
pars,
fun,
residuals
)
# S4 method for nls
rstudents(
model,
varobj,
method = c("Richardson", "simple", "complex"),
residuals = NULL
)
# S4 method for CDFmodel
rstudents(
model,
varobj,
method = c("Richardson", "simple", "complex"),
residuals = NULL
)
# S4 method for nls.lm
rstudents(
model,
varobj,
method = c("Richardson", "simple", "complex"),
fun,
residuals = NULL
)An object from the ""nls" class. It can be derived. e.g., from
function nls or nlsLM from
the R packages stats and minpack.lm, respectively.
Objective variable used build the model.
Argument for grad function from
numDeriv package. One of "Richardson", "simple", or "complex"
indicating the method to use for the approximation.
model parameters.
The named expression (not a character string) of the fitted model. For example, the normal distribution function is given by pnorm (see ?pnorm).
Residuals from the model fitting.
A vector of studentized residuals when model is missing, NULL, or "nls" class object. If \(model="CDFmodel\), then it will update the information carried on model$rstudent and the model will be returned.
If the model argument is provided, then only the argument varobj are required. The internally studentized residuals are computed as \(t = residuals/(s * sqrt(1 - h))\), where \(h\) is the diagonal of the hat matrix and \(s\) is the estimation of the residual variation ((see Wikipedia)).
If the hat matrix cannot be estimated, the standardized residual are estimated as equal to the value of a residual, divided by an estimation of its standard deviation, i.e., \(t = residuals/s\).
Studentized/Standardized residuals greater than 2 and less than -2 are usually considered large.