[R-sig-eco] spatial correlation in mixed-effect models

[aletsia83]

Hello.

Recently, I started to develop regression models using spatial data. In
particular, I've been developing mixed-effect models. My data requires to
take into consideration spatial autocorrelation, so I decided to use the
corStruct function in the nlme library.

Once I built a variogram plot to analyze the autocorrelation, though, the
data (pairs of locations) have a
semivariance (y axis value) that goes above 1. It looks like the corStruct
spatial correlation models are forced to level out at semivariance = 1. This
way, the curve (Gaussian is the closest fit) does not fit my data.

Here is the variogram plot:
http://r-sig-ecology.471788.n2.nabble.com/file/n7577462/gaussian.jpeg 

Does anybody know how to solve this problem?

Hope I'm not misusing the function.
Alessia

--
View this message in context: http://r-sig-ecology.471788.n2.nabble.com/spatial-correlation-in-mixed-effect-models-tp7577462.html
Sent from the r-sig-ecology mailing list archive at Nabble.com.