[sorry if this arrives in duplo: it doesn't show up in the archives and it
seems that the address I posted this from originally is no longer functional]
Hello,
I think I could do with some suggestions concerning the following problem.
I have data from a set of experiments on motion sickness where for each subject,
I have
1) personal data like age and gender
2) a subjective rating of discomfort/sickness, on a 10-level scale,
"sampled" by the subjects as changes occurred. This rating is
normalised per subject to the individual maxima given: Sickness.norm .
3) a conclusive, objective rating sick/not-sick. (WasSick)
4) various objective observables, sampled throughout the experiment. These are
averaged either per time-interval (of say 2") or per "plateau" of
the subjective discomfort rating.
I intend to analyse the averaged observables from (4) (say mean and
power/variance) as a function of the (2), (3) and (1). It would be interesting,
at least for first looks, to be able to do something like
> summary( aov( mean~Sickness.norm*Gender +
Error(Subject/(Sickness.norm*Gender) ) )
or
> summary( aov( mean~Sickness.norm*WasSick +
Error(Subject/(Sickness.norm*WasSick) ) )
To stick with the latter example: my data is sufficiently balanced (about 50%
sick subjects), so I would expect that there are sufficient samples to perform
an ANOVA with. I get the message
"Error in "names<-.default"(`*tmp*`, value = nmstrata) :
names attribute must be the same length as the vector"
I expect that this has to do with the fact that none of my independent variables
are in fact truely independent (controlled by the experimenter), but instead
depend on the subject. For the age, gender and 'WasSick' data, there is
only a single number per subject (indeed, removing them from the anova makes the
analysis "possible").
So I reckon that they must enter elsewhere in the Error() expression...
I'd appreciate it if somebody is capable and willing to explain if and how
this can be done!
Thanks in advance!
R.
