Data files:

stations: names of 82 stations with order numbers
locations: order number, latitude, and longitude
muskrat.dat: orginal muskrat data for all 82 stations (with name of stations and locations)
mink.dat: orginal mink data for all 82 stations (with name of stations and locations)
logmuskrat: log muskrat data for all 82 stations 
logmink: log mink data for all 82 stations 
1r: log muskrat data form the 1st station
4m: log mink data form the 4th station

------------------------------------


From: Nils Christian Stenseth <n.c.stenseth@bio.uio.no>
Subject: muskrat and mink data

Dear Qiwei:-
As an attach file, the muskrat and the mink data follows. They are organized
so that one file contain the muskrat data and another file contains the mink
data; each set cover the same stations. For each file, the data are
organized so that the station name and geographic coordinates are given,
then the 25 years of data terminated by and "end"; then the name, etc for
another station, etc.
Nilschr.
PS: for the Medmonto there seems only to be 16 years worth of data:
I'm trying to sort this out.
PPS: The geographic location is given as follows: the first set of numbers
within the paraenthesis is the latitude, and the second set of numbers are
the longitude; first degrees then minutes (the latidude is North; the
logitude is west and is indicated by a "-"). From this I think it is
possible to find the distances between stations.
.....
Memo
MUSKRAT-MINK MODELLING

Muskrat and Mink data exist for more than 80 individual posts throughout
Canada; geographic locations is known for each of these posts (hence,
distances between these can be obtained). The muskrat and the mink relates
to each other in a prey-predator manner (the muskrat being the prey).

My guess would be that when written out in delaycoordinates, both the
muskrat and the mink corresponds to an order two process; however, this
needs to be tested for.

I've no idea about whether or not these series are linear or not; needs to
be tested for. If non-linear, I'd use a TAR model-approach, possibly a
continuous TAR approach.

Certainly, a bivariate apprach would be applicable; use the modelformulation
described in our (HT/NCS/QW) proposal - here the non-linearities might be
either in the a1/b1 terms or in the a2/b2 terms: viable alternative
hypothesis would be a1/b1 vs a2/b2.

Notice that the muskrat is an ordinary rodent (just like those I've
published extensively on); they have longer cycles than many other rodents,
but that is presumabley due to an habitat effect; do notice that some (very
few) of the series correspond to a 4-year cycle, whereas the majority
correpsond to an appraoximately 10-year cycle. The shorter cycles are in the
tundra whereas the longer are in the boreal forest (where we find the hare
and the lynx).

Very little is published on the muskrat; I have some knowledge about this
species - close colleagues of mine have much more insight on this species
(and its predator).

I would suggest having a preliminary analysis of these data for the purpose
of becomming familiear with them (and the system they represent). Hopefully,
we could privately discuss these preliminary results while together in Oslo
in early May.

Since essentailly nothing has been written on the muskrat-mink system, much
of what can be obtained from the analysis of these data would be of general
inteset: in partciular, the common structure and the estimation of the
parametes under such an assumption - preferably in a bivariate setting with
the above mentioned model - would indeed be of great interest.

I'm prepared to put lots of effort into the formulation of questions and the
interpretation of results refering to these data. Indeed, very intersting stuff.

Nils Chr.
April 18, 1997

With these data, the following issues are of interest: is there any
geogrphic pattern in the preiodicity (I've played a bit on that myself and
there is a tendendy that in the lower western part of the boreal forest
(this will become more clear when you get the map) there is longer periods
than otherwise.) Furthermore, is there any spatial structure in the data -
and even more interesting, is there anything like a wave like movement (this
has been suggested by others).
