> Does not this give you what you need?
> model.tables(rawfixtimedata.aov,"means", se=TRUE)
I tried that, but get an error:
SEs for type 'means' are not yet implemented
Maybe I'm not using the correct terminology to describe what I need to do.
Using the main effect of Marking as an example, I have the following mean
fixation times for each of 12 subjects:
Sub Absent Present
1 1278 586
2 2410 571
3 408 477
4 645 371
5 265 415
6 4871 354
7 1878 790
8 6064 592
9 761 363
10 1073 566
11 1043 383
12 1170 290
The means for markings present and absent, respectively, as reported by
both R and SPSS were:
factor(marking)
Present Absent
480 1822
The standard errors for these means, SE(x) = SD(x)/sqrt(n), should be:
Present Absent
41.42 525.55
Which is what SPSS gives. I need to know how to get R to compute the same
values.
Any suggestions?
Thanks,
Jason
On Dec 13, 2008 1:30pm, David Winsemius wrote:
>
>
> On Dec 13, 2008, at 11:37 AM, Jason Augustyn wrote:
>
>
>
>
> Hi David, thanks for the quick response. I did look at the help files for
model.tables and se.contrast and neither seemed appropriate. I probably
wasn't clear enough in my original email, so here's more information:
>
>
>
> I'm analyzing data from a psychology experiment on how people allocate
visual attention when walking over difficult terrain. In the experiment
subjects walked on a treadmill for 30 minutes while performing an
attention-demanding reaction time task. In one condition they could walk
freely, whereas in another condition they had to avoid markings placed on
the treadmill belt to simulate obstacles. The stimuli for the reaction time
task were placed either at eye-level or near the ground.
>
>
>
> The dependent measure I'm working with comes from an eye-tracking system
that we ran while subjects walked, which provided data on the amount of
time subjects looked at the reaction time stimuli versus the treadmill
belt. We divided the fixation time data sets for each subject into three
time bins to look at changes in fixation behavior over time.
>
>
>
> So the full design of the study is a 2x2x2x3 with repeated measures on
all factors. The factors were:
>
> Markings (Levels: present, absent)
>
> Reaction time task "Position" (Levels: eye-level, ground-level)
>
> Eye fixation "Plane": (Levels: RT stimuli, treadmill belt)
>
> Time bin (Levels: 1,2,3)
>
>
>
> A call to aov yields main effects of Markings, Position, and Plane, as
well as a Markings*Plane interaction. For comparison purposes I ran the
same analysis in SPSS and got equivalent ANOVA results, so I'm confident
the model has been set up properly in R.
>
>
>
> Now, what I want to get are means and standard errors for the main
effects and interaction to generate figures for publication using other
software. As stated in my initial post, I got the means using model.tables
and they are correct as compared with the SPSS output. However, I cannot
get the standard errors for the means. I've tried various things in R and
cannot get values that correspond to the SPSS output.
>
>
>
>
> My understanding is that the se's are for the effects, ie on parameter
estimates for differences, rather than for the means themselves. One get
these (at least in the example on the help page) with:
>
>
>
> model.tables(npk.aov, "means", se = TRUE)
>
>
>
> Does not this give you what you need?
>
>
>
> model.tables(rawfixtimedata.aov,"means", se=TRUE)
>
>
>
> I am not sure what you are referring to when you ask for se's for
the "means" in the presence of interactions. How are you going to partition
the cases? Would one case contribute to both the main "mean" and to any or
all the interaction "means" in which it might be involved?
>
>
>
> --
>
> David Winsemius
>
>
>
>
> Again, I assume there is an R function that can get me the values I need
and would hugely appreciate any pointers. In case you're wondering why I'm
bothering with running the analyses in R given that I already have them
done in SPSS, I'm just generally interested in learning to use R to have an
additional analysis tool in my toolkit.
>
>
>
> Thanks for any help!
>
>
>
> Cheers,
>
>
>
> Jason
>
>
>
>
>
>
>
>
>
> On Sat, Dec 13, 2008 at 9:04 AM, David Winsemius dwinsemius@comcast.net>
wrote:
>
>
>
> On Dec 12, 2008, at 10:59 PM, JS.Augustyn@gmail.com wrote:
>
>
>
> Hi all,
>
>
>
> I'm quite new to R and have a very basic question regarding how one gets
>
> the standard error of the mean for factor levels under aov. I was able to
>
> get the factor level means using:
>
>
>
> summary(print(model.tables(rawfixtimedata.aov,"means"),digits=3)),
>
>
>
> where rawfixtimedata.aov is my aov model. It doesn't appear that there is
>
> an equivalent function to get the standard errors for the factor levels.
>
>
>
> I searched through the help archives and documentation but could not find
>
> anything that would help resolve my problem. I'm sure there is a trivial
>
> solution, but I would sincerely appreciate having someone more expert
>
> dispel my ignorance.
>
>
>
>
>
> Have you looked at the help page for model.tables? ... and
perhaps ?se.contrast
>
>
>
> There are arguments to that function that result in standard errors for
_effects_. If standard errors on the contrasts are not what you wanted,
then perhaps a full example would help.
>
>
>
> --
>
> David Winsemius
>
>
>
>
>
> Cheers,
>
>
>
> Jason Augustyn
>
>
>
> [[alternative HTML version deleted]]
>
>
>
> ______________________________________________
>
> R-help@r-project.org mailing list
>
> https://stat.ethz.ch/mailman/listinfo/r-help
>
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
>
> and provide commented, minimal, self-contained, reproducible code.
>
>
>
>
>
>
>
>
[[alternative HTML version deleted]]