proj
Projection Matrix
Description
Returns a matrix or series of matrices that represent the projection.
This function is an S Version 3 generic (see Methods);
method functions can be written to handle specific
S Version 3 classes of data. Classes which already
have methods for this function include:
aovlist, lm.
Usage
proj(object, ...)
## Default S3 method:
proj(object, onedf = TRUE, ...)
## S3 method for class 'aov':
proj(object, onedf = FALSE, unweighted.scale = FALSE, ...)
## S3 method for class 'aovlist':
proj(object, onedf = FALSE, unweighted.scale = FALSE, ...)
## S3 method for class 'lm':
proj(object, onedf = FALSE, unweighted.scale = FALSE, ...)
Arguments
object |
any object of class lm or any object that inherits from class lm.
For proj.aovlist, the object are objects inheriting from
class aovlist.
It will run faster if the object contains either a qr
component or a proj component.
|
onedf |
logical flag. When TRUE , the function returns a matrix of single-degree-of-freedom
projections of the response variable onto the columns of the
predictor matrix. The default method does not use this argument.
When FALSE, the function
collapses the single degree of freedom projections into multi-degree-of-freedom
projections. Each column of the collapsed result represents one term of the
analysis of variance table. The sum of squares of each column is the sum of
squares for the corresponding term in the model formula, with degrees of
freedom given by the df attribute of the result. The formula itself
is returned in the formula attribute.
|
unweighted.scale |
Whether results are returned on the unweighted scale (useful for
comparing results with the original response) or the weighted scale
(which is the scale used for the data analysis). This only affects
models with weights.
|
... |
any other arguments will be ignored.
|
Details
For proj.default, the object should contain a qr component and a effects component.
For proj.aov, it is not implemented for multiple responses, so object inherits from maov is not supported.
For proj.lm, it is not implemented for mlm filts, so object inherits from mlm is not supported.
Value
matrix of orthogonal columns, one column for each column in an orthogonalized
model matrix. In the default method, each column in the result is the
projection of the response variable onto a column in the Q matrix from the QR
decomposition of the model matrix. The sum across the columns gives the column
of predicted values. The sum of the squared values in each column is the
single-degree-of-freedom sum of squares for the corresponding column of the
model matrix.
If onedf is TRUE, the result matrix with a row for each coefficient,
otherwise, it with a row for each term.
The method for lm objects appends a column of residuals.
See Also
Examples
lm.object <- lm(speed ~ dist, cars)
tmp <- proj(lm.object)