R help - Jun 2010 - Why did TukeyHSD not work when I used it for post-hoc for 2way within-subjects anova?

Dear R people,

I have a couple of questions about post-doc analyses for 2 by 2 within
subjects ANOVA. I conducted a psycholinguistic study that combined a 2 by 2
design and a latin square design. Specifically, I had 32 items each of which
generated 4 conditions. Participants saw each of the 32 items only once: 8
in Condition A, 8 in B, 8 in C, and 8 in D. The table below serves as an
example.

*Subject* *Factor1* *Factor2* *Condition* *ReadingTime* *Item* 1 A A A 238
Item1 1 A A A 369 Item5 1 A A A 494 Item9 *.* *.* *.* *.* *.* *.* *.* *.* *.
* *.* *.* *.* 1 B A C 294 Item22 1 B A C 694 Item26 *.* *.* *.* *.* *.* *.*
*.* *.* *.* *.* *.* *.* 8 A A A 334 Item3 8 A A A 366 Item7 *.* *.* *.* *.*
*.* *.* *.* *.* *.* *.* *.* *.*

I did ANOVA by using the following
:
result.subject=aov(ReadingTime~Factor1*Factor2 +
Error(Subject/(Factor1*Factor2)), data=data), AND
result.item=aov(ReadingTime~Factor1*Factor2 + Error(Item/(Factor1*Factor2)),
data=data).


Both both functions returned significances. Now I am confused as to what
methods I should use to do post-hoc analysis and how to use them. I
attempted TukeyHSD, but it gave me an error message as shown below. I am not
sure how to apply TukeyHSD properly since the help file was not clear enough
on that. So I wonder if you can tell me what went wrong with my function.
> TukeyHSD(result.subject, c('Factor1','Factor2'),
ordered=TRUE)
Error in UseMethod("TukeyHSD") :
  no applicable method for 'TukeyHSD' applied to an object of class
"c('aovlist', 'listof')"

Also, I I attempted *pairwise.t.test* by averaging each participant's
reading times in each condition (see the table below) with *p* values
adjusted with Bonferroni method (*e.g. *pairwise.t.test(y,group,
"bonferroni", pair=TRUE)), . It gave me significant results. But I am
not
sure if it is a proper method to use (Sorry I know this is more of a
statistics related problem). So I wonder if you can tell me if
pairwise.t.test is a proper method. If not, what other functions might be
useful for post-hoc analysis. Thank you in advance!!

*Subject* *Condition* *AvgRT* 1 A 210 2 A 309 . . . . . . 19 A 305 20 A 502
1 B 480 2 B 209 . . . . . . 19 B 400 20 B 270 . . . . . . 19 D 190 20 D 254

	[[alternative HTML version deleted]]

Sorry, that in the last message the tables were messed up. Here is a link to
the tables http://www-scf.usc.edu/~hex/data.txt

<http://www-scf.usc.edu/~hex/data.txt>Thanks!


On Sun, Jun 6, 2010 at 10:18 AM, Xiao He <praguewatermelon@gmail.com>
wrote:
> Dear R people,
>
> I have a couple of questions about post-doc analyses for 2 by 2 within
> subjects ANOVA. I conducted a psycholinguistic study that combined a 2 by 2
> design and a latin square design. Specifically, I had 32 items each of
which
> generated 4 conditions. Participants saw each of the 32 items only once: 8
> in Condition A, 8 in B, 8 in C, and 8 in D. The table below serves as an
> example.
>
> *Subject* *Factor1* *Factor2* *Condition* *ReadingTime* *Item* 1 A A A 238
> Item1 1 A A A 369 Item5 1 A A A 494 Item9 *.* *.* *.* *.* *.* *.* *.* *.*
> *.* *.* *.* *.* 1 B A C 294 Item22 1 B A C 694 Item26 *.* *.* *.* *.* *.*
> *.* *.* *.* *.* *.* *.* *.* 8 A A A 334 Item3 8 A A A 366 Item7 *.* *.* *.
> * *.* *.* *.* *.* *.* *.* *.* *.* *.*
>
> I did ANOVA by using the following
> :
> result.subject=aov(ReadingTime~Factor1*Factor2 +
> Error(Subject/(Factor1*Factor2)), data=data), AND
> result.item=aov(ReadingTime~Factor1*Factor2 +
> Error(Item/(Factor1*Factor2)), data=data).
>
>
> Both both functions returned significances. Now I am confused as to what
> methods I should use to do post-hoc analysis and how to use them. I
> attempted TukeyHSD, but it gave me an error message as shown below. I am
not
> sure how to apply TukeyHSD properly since the help file was not clear
enough
> on that. So I wonder if you can tell me what went wrong with my function.
>
> > TukeyHSD(result.subject, c('Factor1','Factor2'),
ordered=TRUE)
> Error in UseMethod("TukeyHSD") :
>   no applicable method for 'TukeyHSD' applied to an object of class
> "c('aovlist', 'listof')"
>
> Also, I I attempted *pairwise.t.test* by averaging each participant's
> reading times in each condition (see the table below) with *p* values
> adjusted with Bonferroni method (*e.g. *pairwise.t.test(y,group,
> "bonferroni", pair=TRUE)), . It gave me significant results. But
I am not
> sure if it is a proper method to use (Sorry I know this is more of a
> statistics related problem). So I wonder if you can tell me if
> pairwise.t.test is a proper method. If not, what other functions might be
> useful for post-hoc analysis. Thank you in advance!!
>
> *Subject* *Condition* *AvgRT* 1 A 210 2 A 309 . . . . . . 19 A 305 20 A
> 502 1 B 480 2 B 209 . . . . . . 19 B 400 20 B 270 . . . . . . 19 D 190 20 D
> 254
>
>
	[[alternative HTML version deleted]]