Dear all,
I wonder what you think about this. I have data on rabbits released in
artificial warrens (in France rabbits are much appreciated by hunters) and
individually followed at some days. The variable of interest is the
distance from the warren. On those days, observers don't see all rabbits.
Then for each rabbit we have several distances corresponding to the days
when they were observed.
We wan't to study the effect of a rabbit to be released with rabbits of the
same social group ("familiar rabbits") or not ("non-familiar rabbits") upon
the dispersion distance. More particularly, I would like to test whether
the variance of the dispersion distance is greater for non-familiar rabbits
than for familiar ones.
As my observations are not independant (they are grouped according to the
rabbit) I have thought of testing this by bootstrap. As I have 40 rabbits,
I would randomly select 40 rabbits with replacement and one bootstrap file
would be constituted of all observations corresponding to the selected
rabbits. For 20 rabbits in each group, the first 20 rabbits selected would
be attributed the group "familiar" and the last 20 would be attributed the
group "non-familiar". The ratio of the 2 variances calculated in each group
would be computed for each bootstrap file and the observed ratio would be
compared to the computed ratios.
Here are my questions:
- What do you think of randomly selecting rabbits instead of individual
observations? My aim is to preserve the structure of the file.
- Do you think the test procedure is correct? I wan't to estimate the
distribution of the ratio under the null hypothesis (i.e. equality of
variances).
- Do you have other ideas to analyze these data?
Thank you very much in advance for your contribution.
Greetings,
Eve CORDA
Office national de la chasse et de la faune sauvage
5, rue de Saint Thibault
SAINT-BENOIST
78610 AUFFARGIS
BP 20 - 78612 LE PERRAY EN YVELINES Cedex
FRANCE
Tel : +33 (0)1.30.46.60.64
Fax : +33 (0)1.30.46.60.99
Email : [log in to unmask]
|