Yes, it is to compare models, and any ANOVA analysis compares models. What one might normally do, for example,
anova(fit.full)
is a series of model comparisons, specifically the SS of competing models. When only one model fit is supplied, the function attempts to find the series of model comparisons to use; if two or model fits are used, the first one in is considered the null/reduced model for the others. Using your example, shape ~ f1 + f2 + f1:f2, you might get ANOVA tables with effects estimated for f1, f2, and f1:f2. If SS = type I, then here is the series of model comparisons:
effect reduced full
f1 intercept intercept + f1
f2 intercept + f1 intercept + f1 + f2
f1:f2 intercept + f1 + f2 intercept + f1 + f2 + f1:f2
If SS = type 2
effect reduced full
f1 intercept + f2 intercept + f1 + f2
f2 intercept + f1 intercept + f1 + f2
f1:f2 intercept + f1 + f2 intercept + f1 + f2 + f1:f2
This is to say that every line in an ANOVA table is basically an application of anova(reduced model, full model), based on a strategy dictated by the type of SS. The only exception is that the residual SS in the F-statistic uses the full-model residuals.
By using anova(fit.null, fit.full, …), you are directing the analysis to perform a comparison as such:
effect reduced full
f1 + f2 + f1:f2 intercept intercept + f1 + f2 + f1:f2
Here is something else you could do. You could make a new variable,
groups = interaction(f1, f2)
and a new model fit
fit.g <- procD.lm(shape ~ groups, …)
and test this with ANOVA:
anova(fit.g)
and you will get the same exact result as anova(fit.null, fit.full, …) or as anova(fit.null, fit.g) (The effect of group is the same as the combined effects of f1, f2, and f1:f2.) A factorial model can be made into a single-factor model (but as Ian noted, perhaps with some concerns). With a single-factor model, you only have one model comparison to make, so strategies converge.
I think this is the same thing you were asking and I hope my response was not overly pedantic, but I hope this might help anyone reading this thread who is trying to wrap their head around these subtle but similar options.
Cheers!
Mike