NaN Standard Errors for Ecological Distance Covariates in Single- and Multi-Session Models

[Madison Thurber]

Hi all!

I am working with the ecological distance models in oSCR and have run into a few questions regarding model specification and interpretation of the output.

I am working with two study areas (North and South) and am interested in evaluating how different environmental covariates influence movement resistance. I started with forest land cover as my first covariate.

I initially attempted to fit a multi-session ecological distance model using one of my top-ranked models from a previous SCR analysis, with session and standardized forest cover included as covariates for asu.

multimodel3 <- oSCR.fit( model = list( D ~ session, p0 ~ b, sig ~ session + sex, asu ~ session + forest_z - 1 ), scrFrame = scr.multi, ssDF = ss.multi, costDF = cost.multi, distmet = "ecol", PROJ = "+proj=utm +zone=17 +datum=NAD83 +units=m +no_defs", plotit = FALSE, trimS = 8 )

However, I received the following error:

[1] "session" Error in oSCR.fit(...): I cant find theses covariates in 'costDF'

To work around this, I created session-specific forest covariates (forest_north and forest_south) and fitted the following model:

multimodel3 <- oSCR.fit( model = list( D ~ session, p0 ~ b, sig ~ session + sex, asu ~ forest_north + forest_south - 1 ), scrFrame = scr.multi, ssDF = ss.multi, costDF = cost.multi, distmet = "ecol", PROJ = "+proj=utm +zone=17 +datum=NAD83 +units=m +no_defs", plotit = FALSE, trimS = 8 )

The model completed successfully (AIC = 1950.852), but I noticed that the standard error for c.beta.forest_north was returned as NaN.

Estimate SE z P(>|z|) c.beta.forest_north -0.051 NaN NaN NaN c.beta.forest_south -0.547 0.127 -4.308 0.000

I was concerned about the missing standard error, so I also fitted a single-session ecological distance model using only the North study area to see whether the issue persisted.

northmodel2 <- oSCR.fit( model = list( D ~ 1, p0 ~ t + b, sig ~ sex, asu ~ forest_z - 1 ), scrFrame = scr.north, ssDF = ss.north, costDF = cost.north.single, distmet = "ecol", PROJ = "+proj=utm +zone=17 +datum=NAD83 +units=m +no_defs", plotit = FALSE, trimS = 8 )

This model also completed successfully (AIC = 861.2548), but again returned a NaN standard error for the forest covariate.

Estimate SE z P(>|z|) c.beta.forest_z -0.015 NaN NaN NaN

I have a few questions that I was hoping someone might be able to help with:

1. Is receiving a NaN standard error for an ecological distance covariate something to be concerned about, even when the model completes successfully? Could this indicate an issue with parameter estimation or the covariate itself?

2. For multi-session ecological distance models, is using separate session-specific forest covariates in costDF the appropriate approach, or is there a way to include a session effect directly in the asu formula?

I am relatively new to ecological distance modeling, so any guidance on model specification or interpreting these results would be greatly appreciated!

Thank you in advance for your help!

Madison