Replies: 1 comment 1 reply
-
The most important aspect is that effects that are linear and act in an additive way in log-scale (i.e. the linear predictor) are "log-linear" and multiplicative on the "natural" scale. I.e. if the elevation coefficient is So there's no "free lunch" here; either one deals with the raw coefficients, the multiplicative change, or the derivative at a reference level when keeping all else fixed, where the third option gives a different value for each reference level and values of the rest of the model. The "multiplicative change" interpretation is probably the clearest, as it describes the ratio between two density values at a unit difference in the covariate, keeping all else fixed, regardless of what the rest of the model is fixed to. |
Beta Was this translation helpful? Give feedback.
-
Hi,
I wonder if you're able to provide some help on how to correctly backtransform LGCP model outputs?
Specifically, if you wanted to outline how density varies in relation to covariates within the model, what would be the correct transformation to describe the fixed effect estimates - ie. As distance increases by 1km, density increase by X individuals/km, or similar?
For example, the tutorial below is helpful in plotting how elevation affects log(density), but would be really helpful to understand how you could backtransform the estimates to produce a snappy summary of the relationship?
https://inlabru-org.github.io/inlabru/articles/2d_lgcp_covars.html
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions