Hello, I have been browsing this tutorial on SDM modelling with terra.
Thanks for the awesome tutorial. I have learned a lot from it as I have just started to explore SDMs.
But I have I noticed something that may be important, I understand the page is under construction but it may help in the future.
You model presence absence data (1 and 0's) with a glm with a Gaussian distribution. Wouldn't it be more adequate to use a model with 'family = binomial()' such as:
m1 <- glm(pb ~ bio1 + bio5 + bio12, data=sdmdata, family = binomial())
I have tried both and the predicted values a more aligned to "suitability" values from 0 to 1.
Anyway, thanks for you time, and thanks for the continuous support to the research community!
Cheers,
@francisvolh
Hello, I have been browsing this tutorial on SDM modelling with terra.
Thanks for the awesome tutorial. I have learned a lot from it as I have just started to explore SDMs.
But I have I noticed something that may be important, I understand the page is under construction but it may help in the future.
You model presence absence data (1 and 0's) with a glm with a Gaussian distribution. Wouldn't it be more adequate to use a model with 'family = binomial()' such as:
m1 <- glm(pb ~ bio1 + bio5 + bio12, data=sdmdata, family = binomial())I have tried both and the predicted values a more aligned to "suitability" values from 0 to 1.
Anyway, thanks for you time, and thanks for the continuous support to the research community!
Cheers,
@francisvolh