Xiao He
2010-Jun-10 05:04 UTC
[R] Help with Post-Hoc tests for TWO-WAY within subject ANOVA
Dear R users, I posted a couple of questions and got no response, so I am giving it another shot. I ran an experiment with a TWO-WAY within subject design. A sample dataset is in http://www-scf.usc.edu/~hex/data.txt I already ran ANOVA by using the following formula: aov(RT~Factor1*Factor2 + Error(Subject/(Factor1*Factor2)), data=data) and I obtained the following information: ------------- Error: Subject Df Sum Sq Mean Sq F value Pr(>F) Residuals 19 6154709 323932 Error: Subject:Factor1 Df Sum Sq Mean Sq F value Pr(>F) Factor1 1 131017 131017 1.2624 0.2752 Residuals 19 1971836 103781 Error: Subject:Factor2 Df Sum Sq Mean Sq F value Pr(>F) Factor2 1 48042 48042 0.4281 0.5207 Residuals 19 2132016 112211 Error: Subject:Factor1:Factor2 Df Sum Sq Mean Sq F value Pr(>F) Factor1:Factor2 1 500137 500137 7.3702 0.01374 * Residuals 19 1289336 67860 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Error: Within Df Sum Sq Mean Sq F value Pr(>F) Residuals 560 54063256 96542 ---------------- Now my question is, what kinds of post-hoc tests can be used to examine which of the 4 conditions generated by the 2X2 design is different. From the research I did online, I found that TukeyHSD cannot be used because the aov() function I used contains random effects. I also read that pairwise.t.test adjusted with the Bonferroni method is not good either because it assumes that the null hypothesis is not rejected, hence overcorrecting TYPE 1 error. It would be great if someone could shed some light on my problem. Thank you in advance! Xiao [[alternative HTML version deleted]]