I have a question about data analysis for the list:

I'm working on a couple of endangered grouse species (capercaillie and =
black grouse) in Scotland.
Both species have declined markedly in over the last few decades and =
although the data I have available isn't comprehensive, it is =
potentially very valuable.

I have data on breeding success for both species that goes back as far =
as 1989 from 11 different sites in northern Scotland. However, only two =
different sites have the complete sequence of annual data.

I'm interested in examining the effects of weather patterns and habitat =
on breeding success of the two species.=20

Previous work has suggested that weather patterns during Spring are =
important.=20

I therefore have temperature and rainfall data for Spring for all years =
since 1989. However, weather data is not site specific and comes from =
the nearest weather station. This essentially means that sites in one =
catchment share the same weather data, which is then different from the =
weather data shared by the remaining sites in a second catchment.

I'm also interested in the interaction between weather patterns and =
habitat quality. However, I only have comprehensive habitat data from =
1998. I intend therefore to assume that habitat hasn't changed markedly =
across the years.


To examine the effects of weather conditions, my initial approach has =
been to use multiple linear regression (the variables are either normal =
or readily transformed) with breeding success as my dependent variable =
and the weather variables as my independent variables. My problem is =
that over the 15 year period (1989 to 2004), the weather station data =
shows that spring has become significantly warmer. Over the same period, =
breeding success for the birds has declined. I therefore end up with a =
model showing that colder spring temperatures are better for breeding =
success, which works fine as a description of what has happened but =
doesn't make biological sense.  How do I take account of the spurious =
pattern in the data in my model building? Essentially, if you like, I =
wish to account for the effects of the changes in climate by year before =
examining the effects of weather patterns.

Additionally, is it sensible to examine any interactions between weather =
and habitat variables given that there is no between year habitat =
variation and no between site weather variation?=20


I'm working in R, so any specific help relating to that package would =
also be welcome.

Many thanks,

Martin Dallimer


Dr Martin Dallimer
Woodland Grouse Scientist
The Game Conservancy Trust
Drumochter Lodge
Dalwhinnie
Inverness-shire
PH19 1AF
Scotland
Phone: 01528 522300
Mobile: 07734 537654
email: [EMAIL PROTECTED]

Reply via email to