Update a fac.aov Model
DESCRIPTION:
Updates a fac.aov model. In the special case of reduced models of an orthogonal saturated model, this retains the saturated effects and corresponding error estimates.
USAGE: 
update.fac.aov(object, formula.new, ...)
REQUIRED ARGUMENTS:
object:
a fac.aov object.
formula.new:
a formula describing the new model.
OPTIONAL ARGUMENTS:
...:
arguments passed to update.default.
VALUE:
a fac.aov object, fitting the formula given.
DETAILS:
In the case where object is a saturated model (object$df.residual == 0) from an orthogonal design matrix, and formula specifies a true submodel of object, update keeps the effects from the full model of object, along with estimates of pse, tse and ase. Thus, the reduced model retains estimates from the full model for use in functions that examine the model, e.g., pareto, summary. However by default, the mse of the reduced model is used for significance tests, unless otherwise specified.

In all other cases, the value returned by update is the same as that returned by fac.aov.

NOTE:
In the reduced model case, the values returned by fac.aov and update, for the same fitted model will have different values for pse, tse, ase and seffects. Estimates in both cases are valid, however those from the update function are based on the effects of the saturated model and are more easily interpretable.
SEE ALSO:
update.formula , fac.aov , summary.fac.aov , plot.fac.aov .
EXAMPLES: 
buffer.fac <- fac.aov(buffer.df)
summary(buffer.fac)
pareto(buffer.fac)


# now compare the results of update and fac.aov
# for fitting a reduced model
buffer.rmod <- update(buffer.fac, ~ pH*thimer + pH*gent)
summary(buffer.rmod)
pareto(buffer.rmod)
buffer.fac2 <- fac.aov(rate~pH*thimer+pH*gent,buffer.df)
summary(buffer.fac2)
pareto(buffer.fac2)


# look at diagnostic plots for the reduced model
plot(buffer.rmod)


# refit using a transformation of the response
buffer.sqrt <- update(buffer.fac, sqrt(rate) ~ .)