Maximilian Pichler

Hi @YJ781, ### Over/underdispersion ASVs == Species, right? Then the under/overdispersion isn't connected to the number of ASVs, because it is ASV/species specific. Let's take a look at the first...

Hi @Zurapiti, In general this is a valid approach, you correct for spatial/temporal trends + autocorrelation by modeling all your spatial and temporal variables with a DNN, which is perfectly...

Hi Jazz, Normally, the number of species does not have a large effect on numerical stability. However, it may happen that the distribution of species is so skewed (dispersion problems)...

@Zurapiti @metagenAu new version with negative Binomial is on the way to CRAN

@metagenAu , @Zurapiti Let me know what you think, it is still experimental and I am not sure how stable it is. If you run into NAs during training, restart...

@metagenAu I would use it in the object where you set the intercept - so in the env object. Technically it doesn't matter, it will 'only' affect your anova (if...

Yes, I know, they could be estimated via MLE, but a reliable implementation (so that it works hand in hand with stochastic gradient descent) requires a lot of time and...

Hi @agnes-duhamet , You can avoid the errors by scaling your variables: ``` library(sjSDM) library(tidyverse) model % scale, formula = ~dist_land+depth), spatial = linear(data = SP %>% scale, formula =...

Hi @agnes-duhamet, No it is possible, the problem is that you have empty levels in your habitat_principal variable: ``` library(sjSDM) library(tidyverse) env_var_scaled = env_var %>% mutate(dist_land = scale(dist_land), depth =...

I think, that's the wrong link?^^ However, what shall we use as default plot? We need a plot.sjSDM method. A heatmap?