[R] trouble with model.tables SE means

[Jon Prince]

David Winsemius wrote:
>
>
> On Dec 22, 2009, at 5:19 PM, Jon Prince wrote:
>>
>> David Winsemius wrote:
>>>
>>> On Dec 22, 2009, at 4:22 PM, Jon Prince wrote:
>>>
>>>> Hi, I'm new to R, with some experience with Matlab and SPSS. I've 
>>>> figured out how to run my repeated measures anova and am getting 
>>>> the right numbers for my effects (comparing with results from other 
>>>> software), but am having trouble with the model.tables function. 
>>>> Specifically, using:
>>>>
>>>> prints the means, but then won't do the SE values, instead giving:
>>>>
>>>>  Warning message:
>>>> In model.tables.aovlist(fm, "means", se = TRUE) :
>>>> SEs for type 'means' are not yet implemented"
>>>>
>>>> Asking for SEs for "effects" works fine, but that's not what I 
>>>> want. I searched the help for this issue and one other person has 
>>>> had this problem last year 
>>>> (http://markmail.org/message/k5yxxqcfiihvzvtp?q=list:r-project+model%2Etables), 
>>>> but the person helping them was unable to replicate it, inferring 
>>>> that it was an out-of-date version. My version is:
>>>>
>>>>  R version 2.10.1 (2009-12-14)
>>>>
>>>> I only downloaded it the other day, and therefore cannot have an 
>>>> outdated version. How can I fix this error and get my SE values? 
>>>> Apologies if I have not provided sufficient information, and thanks 
>>>> in advance for your help.
>>>
>>> When I look at the output of the first model.tables call copied from 
>>> the help page, I see a list element that holds "se" values. Try:
>>>
>>> model.tables(fm,"means",se=TRUE)$se
>>>
>>
>> Thanks for the rapid reply! Unfortunately adding the $se returns 
>> NULL, and repeats the same warning message ("...not yet 
>> implemented"). If you're not experiencing the issue, is it possible 
>> for me to replace the relevant code/source file with what you have 
>> (or would that require recompiling)? Could this be an OS issue? I'm 
>> running Mac OSX 10.6.2.
>>
>> By the way, I "replied all" on this message, but let me know if that 
>> is not the preferred convention. Cheers,
>
> Reply all. That way people can correct my mistakes and general 
> cluelessness. I'm running MacOSX 10.5.8 so it would seem less likely 
> that is the explanation.
>
> 1) Did you run the example in the help pages?
>
> 2) When I look at :
>
> > methods(model.tables)
> [1] model.tables.aov*     model.tables.aovlist*
>
> ... I see both an "aov" method and an "aovlist" method. Is it possible 
> that there is something about the object that you are working on that 
> makes it an aovlist at thus invokes a different function than what the 
> help page invokes?
>
> The code would not require complination... it's available with 
> getAnywhere() and I do not see any calls to compiled or .Internal 
> subroutines. Tell me what happens with the above questions first.
>

Sorry for the delayed response, I've been trying to work out the 
discrepancy between my data and the example data in terms of getting the 
SE means.

1.) Upon trying the example, it appears to work on my machine (i.e., I 
get the SE means). Since then I've been spending my time trying to 
figure out why it doesn't do the same for my data. I can't figure it out.

2.) Given my lack of experience, it is quite possible that I've loused 
up an object somewhere, but I don't know how to track that down. I have 
put my code below, and can send the data file too if you're willing to 
take a look. Both the p and t factors have three levels, combined 
factorially. There are four replications of each p-t combination, and 
the rating data vary from 1 to 7.

##load data
    rate_data<-read.table("/Users/jonprince/Desktop/attend_pitch_3p3t.txt")
##set factors
    p<-factor(rate_data$V3)
    t<-factor(rate_data$V2)
    subject<-factor(rate_data$V1)
##set data
    rate<-rate_data$V4
##run anova
    fm<-aov(rate ~ p*t + Error(subject/(p*t)),rate_data)
##get summary
    summary(fm)
##tables
    model.tables(fm,"means",se=TRUE)

Thanks again, I really appreciate the help.

Jon


-- 
Jon Prince
Postdoctoral Research Associate