|
|
Bethwell,
>>I want to determine differences in activity budgets of some animal species
between different seasons. Activities were recorded as either feeding,
walking, resting. Observations were after every 30 minutes from 0800 - 1600
hrs. I want to present the data as percentage time spent feeding,
walking..etc. If out of 16 observations, feeding was 12 observations. Is it
proper to say 75% was spent feeding. Or do I need to assume one observation
represents 30 minutes.
How you characterize the result depends, I think, on how the observations
were made. Suppose that every 30 minutes you looked at an animal for two
seconds and recorded what it was doing at that two second moment, which was
then recorded as what the animal was doing for the 30 minute block. Versus,
suppose you watched an animal continuously for 30 minutes, recorded how much
time it spent doing each activity and recorded as the activity for the 30
minute block the activity having the longest total duration.
>>Can I run an analysis of variance to compare differences in activities
between seasons.
I think that you can not/should not run a simple anova for the following
reasons. 1) you have repeated measures. 2) You have outcomes that are
proportions. 3) Your three activity variables are exhaustive and sum to
1.00. On 1) you observe a subject animal across multiple seasons and each
animal has three activity measures. On 2) the variance of a proportion is a
function of the proportion, (p*(1-p)) and anova assumptions (variance
homogeneity) will be violated. An arcsine transformation is commonly used to
fix this problem (but google arcsine transform and look at the first five or
six nonsponsored results because the issue seems to be a bit more
complicated than just do an an arcsine transform). I don't have enough
knowledge to say how an arcsine transform works in a repeated measures
design. There are other more knowledgeable people on the the list. On 3) I
think that having a set of variables that sums to 1.0 will cause problems.
Perhaps an arcsine transform will fix this problem also. I defer, however,
to more knowledgable responders. Perhaps some references will be suggested.
Gene Maguin
|
Locked
