Cross Validated
2026-11-01 14:15 UTC
By Tanya Lemieux
AI-113-20261101-social-media-84cb6ff5
Not sure if I have complete separation in binomial GLMM and what to do
I used acoustic telemetry to gather fish habitat use data that I am now trying to run a glmm analysis on. I converted the acoustic detection data into hourly presence/absence data for a single location - an off channel pond to a river system. The reason for doing this is because I want to know what factors influence fish use of that specific pond. This means I can't use location as a random effect since I only have one but I must have time as a random effect in some fashion because I repeatedly sampled that single location. I also need to have individual fish ( transmitter_id ) as a random effect since I also repeatedly sampled the same fish. The best model I have found so far (i.e., passes most DHARMa tests, has lowest AIC, and has best z scores) is: np_bscp_mod7 total_length_mm is numerical and is the total length of each fish in mm. mu_binary is a binary variable with 1 indicating muskellunge present in the habitat and 0 indicating no muskellunge present in the habitat (this excludes the fish in the row this data is added to). np_binary is the same but refers to northern pike. wa_binary is the same but refers to walleye. ws_binary is the same but refers to white sucker. Temperature is numerical and is the temperature within the pond. avg_discharge_m3s is numerical and is the discharge in the river. DO_mgl is numerical and is dissolved oxygen in the pond. study_hour is a factor and is the hour of day. I added the times column to account for the temporal pseudoreplication a…
I used acoustic telemetry to gather fish habitat use data that I am now trying to run a glmm analysis on. I converted the acoustic detection data into hourly presence/absence data for a single location - an off channel pond to a river system. The reason for doing this is because I want to know what factors influence fish use of that specific pond. This means I can't use location as a random effect since I only have one but I must have time as a random effect in some fashion because I repeatedly sampled that single location. I also need to have individual fish ( transmitter_id ) as a random effect since I also repeatedly sampled the same fish. The best model I have found so far (i.e., passes most DHARMa tests, has lowest AIC, and has best z scores) is: np_bscp_mod7 total_length_mm is numerical and is the total length of each fish in mm. mu_binary is a binary variable with 1 indicating muskellunge present in the habitat and 0 indicating no muskellunge present in the habitat (this excludes the fish in the row this data is added to). np_binary is the same but refers to northern pike. wa_binary is the same but refers to walleye. ws_binary is the same but refers to white sucker. Temperature is numerical and is the temperature within the pond. avg_discharge_m3s is numerical and is the discharge in the river. DO_mgl is numerical and is dissolved oxygen in the pond. study_hour is a factor and is the hour of day. I added the times column to account for the temporal pseudoreplication a…
Full article content could not be extracted automatically. Read the original below.
Source:
Cross Validated
· stats.stackexchange.com