Hello,
I am analyzing time-series data for multiple songbird species in northern
Canada where data were collected at 3 point count stations within a stand
(~150 stations) visited twice a year and with multiple observers. I am using
a linear mixed effects model (lme4) that includes year as a fixed effect and
observer, station nested within stand (to account for spatial
auto-correlation) and visit as random variables:
lme3<-lmer(Abundance ~ Year + (1|Site/Station)+ (1|Observer)+ (1|Visit),
data=AMRE, family = poisson(link=log))
To test the model, I simulated a data set (Poisson distribution) with a 50%
decline over 50 years (with a mean abundance at year 1 of 10 individuals of
species X).
The parameter estimates for the intercept and year are:
Estimate Std. Error t value
(Intercept) 2.33181 0.03654 63.82
Yr -0.17089 0.01685 -10.14
Since the data are log transformed during the analysis (link=log), I have to
apply an anti-log to the parameter estimates to get the 'true' value
(i.e.,
exp(2.33181) = 10.29656 for the estimate of abundance and
exp(-0.17089)0.8429143, so 1-0.8429143= 0.1570587, for the slope from the year 1
to year
2). Assuming that the method of back-transformation is correct, how do I
calculate the slope for the subsequent time steps (e.g., year 2 to year 3)
and for the overall trend (year 1 to year 5) using the parameter estimate
seeing as the slope should be ~ -1.25.
Your help is appreciated!
Kevin Kardynal, MSc
Environment Canada
--
View this message in context:
http://n4.nabble.com/Calculation-of-slope-for-Poisson-regression-tp961250p961250.html
Sent from the R help mailing list archive at Nabble.com.
