coefficients
Extract Information from a Model
Description
Functions that extract the coefficients, residuals, and fitted values
from a fitted model.
Usage
coefficients(object, ...)
coef(object, ...)
## Default S3 method:
coef(object, ...)
## S3 method for class 'listof':
coef(object, ...)
residuals(object, ...)
resid(object, ...)
fitted.values(object, ...)
fitted(object, ...)
Arguments
object |
any object that represents a fitted model, or, by default any object
with a component named by the name of the extractor function.
|
... |
additional arguments to pass to the function defined in the object.
|
Value
returns the the coefficients, residuals, or fitted values defined by
the model in object.
While for some models the return values are identical to the component
of the object with the same name, we recommend that you use the
extractor functions, since these will call the appropriate method for
this class of object.
For example, residuals from generalized linear models come in four
flavors, and the most useful one is typically not the component.
Methods
The following table displays the available methods for each function.
Function: | "coef" | "residuals" | "fitted" |
aov | x | | |
arima | x | x | x |
default | | x | x |
discrim | x | | |
factanal | | x | x |
glm | | x | |
listof | x | | |
lm | x | x | |
lmList | x | x | x |
nls | x | x | x |
|
Note: To encourage you to use the extractor function rather than the
component, there are three abbreviated versions of these functions:
coef();
resid(); and
fitted().
Singular models
When a model is over-specified, the standard fitting method permutes
the columns of the model matrix, which therefore also permutes the
corresponding coefficients and effects.
This is another case where the components for these quantities behave
a little differently from the values of the corresponding functions.
If
fit is a fitted, over-determined model, then
coef(fit) only returns the full-rank part of the coefficients,
while
fit\$coef returns coefficients for all the columns in the
original model matrix.
In the latter case, elements that are aliased with earlier degrees of
freedom are returned as missing values (
NAs).
In breaking down the coefficients or effects by terms, the distinction is also present.
The component
assign refers to the original model matrix, so it
defines the degrees of freedom in
fit\$coef.
Another component,
R.assign is produced in the singular case to
index terms in the
R matrix, and also in
coef(fit).
For more details, see
lm.object.
See Also
Examples
kyph.glm1 <- glm(Kyphosis ~ Age + Start + Number, family=binomial,
data=Sdatasets::kyphosis)
coef(kyph.glm1)
## (Intercept) Age Start Number
## -2.036932 0.01093048 -0.20651 0.410601
residuals(kyph.glm1)
## 1 2 3 4 5
## -0.7707923 -0.5111615 1.189314 -1.106713 -0.2461915
## .
## .
## .