lm
,
aovlist
dummy.coef(object)
lm
or
aovlist
, or one that inherits from these classes.
The coefficients returned by
coefficients(object)
for factor terms in
an
lm
object are coefficients for the contrast variables derived for those factors.
By default these are helmert contrasts for unordered factors and polynomial for ordered.
For a main effect term, there are typically
K-1
coefficients for a
K
level factor.
dummy.coef
returns
K
coefficients for a
K
level factor, and these coefficients will satisfy the constraints implied by the choice of contrasts.
For example if the default or
"contr.sum"
contrast functions are used to code a
K
-level factor,
K
coefficients will be returned by
dummy.coef
for this term, and if there was an intercept in the model, they will
sum to 0.
Similarly
K . L
coefficients are returned for an interaction term between a
K
level and a
L
level factor, with the appropriate constraints honored.
Coefficients for numeric variables and matrices are returned as well, but these are the same as those on object, unless they are crossed with factor terms.
myfit <- lm(Yield ~ Temp + Cat + Temp:Cat, data = catalyst) myfit # produces the following output: # Call: # lm(formula = Yield ~ Temp + Cat + Temp:Cat, data = catalyst) # Coefficients: # (Intercept) Temp Cat Temp:Cat # 64.25 11.5 0.75 5 # Degrees of freedom: 8 total; 4 residual # Residual standard error: 3.708099 dummy.coef(myfit) # produces the following output: # $"(Intercept)": # (Intercept) # 64.25 # $Temp: # 160 180 # -11.5 11.5 # $Cat: # A B # -0.75 0.75 # $"Temp:Cat": # 160A 180A 160B 180B # 5 -5 -5 5