Date as a Covariate

joel

New Member
#1
Hi, I am looking at the effect of various weather variables on sub-population size of a certain species of snake. Sub-population size at the study site over the active period follows a quadratic shape increasing in the spring to a peak in the summer and drops in the fall. Temperature during that period follows a similar pattern. I want to determine the effect of mean weekly temperature, wind speed, and precipitation on weekly sub-population size. A quadratic regression of sub-population size vs weather would show a significant effect of temperature. However, shouldn't I include a covariate to account for the seasonal variation? The relationship between sub-population size and time is not linear so I have tried using the polynomial 'week^2' as the time covariate but of course this causes the effect of temperature to disappear.

How do I tease apart the effect of temperature and basic seasonal behaviour? Is this possible? Any help on this matter would be greatly appreciated as I have spent many fruitless hours searching in the literature for a suitable solution. Thank you.
 

maartenbuis

TS Contributor
#2
The first question would be: what is temperature nett of season and is its effect the effect you are looking for? I can imagine that the effect of deviations in temperature from season average could sometimes be meaningful, but I doubt whether that is always the case.
 

noetsi

Fortran must die
#3
Seasonal variation in what? You should decide what is actually varying (say sunlight or wind or whatever) and specifically model that.
 
#4
beware of ecological trends
... not "ecological" in the sense that something is a study in the field of Ecology, but trends of two things that happen to occur in a similar seasonality pattern