[R-sig-ME] factors with multiple levels

Martin Maechler maechler at stat.math.ethz.ch
Mon Feb 16 12:12:16 CET 2009

>>>>> "MY" == Miao-Hsuan Yen <melissayen at gmail.com>
>>>>>     on Mon, 16 Feb 2009 17:46:17 +0800 writes:

    MY> Dear lmer list I have one within-subject factor with 3
    MY> levels (A1 A2 A3). By default, I will get an intercept
    MY> for the reference group (A3) and two estimates for A1
    MY> and A2 compared to A3. How can I get an estimate for the
    MY> main effect of factor A as the traditional ANOVA does?
    MY> Is there any function or option that allows me to have
    MY> an overall test?

    MY> In addition, if I have two within-subject factors (A1 A2
    MY> A3 and B1 B2) and their interactions, I will get 2
    MY> estimates for factor A, 1 estimate for factor B and 2
    MY> estimates for the interaction. A similar question is,
    MY> how can I get an estimate for the interaction as the
    MY> ANOVA does?

To both questions:

You compare the model *with* the corresponding factor (or interaction)
term with a model *without* the corresponding term.
That is *the* (generalization of the) classical ANOVA,
and works for many kinds of models in  S and R  via

anova(<model1>,  <model2>, ..)

--> the beginning of 'Examples' on    help(lmer)  aka  ?lmer
does exactly do this ..

    MY> Thanks a lot.

you're welcome;
Martin Maechler, ETH Zurich

    MY> Best, Miao-Hsuan Yen

    MY> 	[[alternative HTML version deleted]]

    MY> _______________________________________________
    MY> R-sig-mixed-models at r-project.org mailing list
    MY> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

More information about the R-sig-mixed-models mailing list