Dear all
I try to do a simultaneous autoregressive (SAR) models to take the spatial correlation into account (Kissling & Carl, 2008). I already selected the best combinations of weight list and neighbours distance by using the value of AIC and minRSA. But I have problem on how to calculate the relative importance of each predictor, especially on calculating the model fit using pseudo-R2 (Pearson correlation of observed and predicted values) and Akaike weights (w).
a method in Rachael V. Gallagher Journal of Biogeography (J. Biogeogr.) (2016) 43, 1287–1298
Can anyone give some suggestions on how to do the above analysis?
Thanks
To give a appropriate fine-scale spatial structure, see page 54 in this link below. You can use free software too.
https://www.ecoevol.ufg.br/sam/SAMTutorial.pdf
https://www.ecoevol.ufg.br/sam/
Calculating the relative importance of any AIC value can be negative (AIC < 0). Actually, the absolute values of the AIC scores do not matter. In practice, criteria of lower AIC (-243) is preferred over the model with AIC = 67.
It is proper that sometimes values became negative A.I.C. However, fitting all models to the same data can lead to "negative log-likelihood density function", especially data with small sample size, and relatively large number of parameters.,
Hope this helps
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis