[R-lang] assessing 3-way factors and their interactions
Nina Kazanina
nina.kazanina@bristol.ac.uk
Fri May 14 07:33:06 PDT 2010
Hello R-lang,
I am analysing an experiment which essentially followed a 2x3 design with
factors Prime (related/unrelated) and Condition (conA/conB/conC). The
dependent variable is Response Time (RT) and I model it by fitting linear
mixed-effect models to the data using 'lmer'.
I am a little lost on how to assess the main effect of Condition and the
Prime x Condition interaction, and in particular whether the different
methods (a), (b) and (c) listed below are each valid and comparable to each
other.
1. Assessing a 3-level factor Condition(conA/conB/conC) as a whole:
(1a) Report the results of a chi-square test from anova(M1, M2): [assuming
that Prime is a significant predictor]
M1: RT ~ Prime + (1|subj) + (1|item)
M2: RT ~ Prime + Condition + (1|subj) + (1|item)
(1b) Report F-value and p(MCMC) from the function aovlmer.fnc(M2, x$mcmc,
c("conB", "conC")). If this option is correct, does it matter whether
effect-coding or treatment-coding of factors Prime and Condition is used?
2. Assessing the interaction Prime x Condition:
(2a) Reporting the results of a chi-square test from anova(M2, M3):
M2: RT ~ Prime + Condition + (1|subj) + (1|item)
M3: RT ~ Prime * Condition + (1|subj) + (1|item)
(2b) Report F-value and p(MCMC) from the function aovlmer.fnc(M2, x$mcmc,
c("Primeunrelated:ConditionconB", "Primeunrelated:ConditionconC")). Again,
as in 1(a) above, does it matter whether effect-coding or treatment-coding
of factors Prime and Condition is used?
(2c) Use effect coding and report the t-value and the p(MCMC) value for the
interaction term Prime x Condition from the model M3
I'd be grateful if someone directed me to a paper that had a similar design,
uses linear-mixed effects model and reports it in a nice clear way.
Many thanks in advance!
Nina
------------------------
Nina Kazanina
Department of Experimental Psychology
University of Bristol
12a Priory Rd.
Bristol, BS8 1TU, UK
Phone: +44 (0) 117 92 88551
Email: nina.kazanina@bristol.ac.uk
Web: http://nk.psy.bris.ac.uk/
------------------------
Nina Kazanina
Department of Experimental Psychology
University of Bristol
12a Priory Rd.
Bristol, BS8 1TU, UK
Phone: +44 (0) 117 92 88551
Email: nina.kazanina@bristol.ac.uk
Web: http://nk.psy.bris.ac.uk/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.ucsd.edu/mailman/private/ling-r-lang-l/attachments/20100514/29423634/attachment.html
More information about the ling-r-lang-L
mailing list