Least Median of Squares Object

DESCRIPTION:

These are objects of class lms They represent a robust fit of a linear model.

GENERATION:

This class of objects is returned from the lmsreg function.

METHODS:

The "lms" class of objects has methods for the following generic functions:

plot, print, summary.

STRUCTURE:

The following components must be included in a legitimate lms object.

VALUE:

coefficients
a vector or matrix of coefficient estimates. This is a matrix when there is more than one response variable; columns represent the response variables, rows represent the explanatory variables.
scale
residual scale estimate for each column of y.
rsquared
robust version of R squared for each column of y. This is 1 minus the square of the quantity: the quanth-order statistic of the absolute value of the residuals over the quanth-order statistic of the absolute value of y - theta, where theta is minimized over the denominator. Here quan is the order statistic to be minimized, and theta is not subtracted from y if intercept = FALSE.
objective
the minimum value found of the quanth-order statistic of each column of the absolute value of the residuals, where quan is the order statistic that is being minimized.
residuals
object like y containing the residuals from the regression. Rows corresponding to missing values in x or y have NAs.
lms.wt
object like y of weights for use in weighted least squares. These weights are 1 for points with reasonably small residuals and are 0 for points with large residuals. This is only included if wt is TRUE in the call to lmsreg.
intercept
same as the input or default intercept.
diagnostic
object like y of resistant diagnostics, For each trial the absolute value of the residuals is divided by the objective for that trial; the maximum of these numbers is retained for each observation. The diagnostic component consists of these maxima divided by their median for each response variable. This is only included if diagnostic is TRUE in the call to lmsreg. If diagnostic=FALSE, then computation as well as space is saved.
method
character string giving the method (Least Median of Squares), the number of subsamples, the size of the subsamples, and the number of subsamples leading to a singular system of equations.
RD
optionally (if mve=TRUE), the robust distances of the cases obtained from cov.mve are returned.
X
same as the input x in a call to lmsreg.default or the data matrix.
model
optionally the model frame, if model=TRUE (in lmsreg.formula).
x
optionally the model matrix, if x=TRUE (in lmsreg.formula).
y
optionally the response, if y=TRUE (in lmsreg.formula).

SEE ALSO:

, , , .