A brief introduction to mixed effects modelling and multi-model inference in ecology

Harrison, Xavier A.; Donaldson, Lynda; Correa-Cano, M.; Evans, Julian C.; Fisher, David N.; Goodwin, Cecily E. D.; Robinson, Beth S.; Hodgson, David J.; Inger, Richard

doi:10.7717/peerj.4794

Cited by 1,472 publications

(956 citation statements)

References 92 publications

Supporting

Mentioning

874

Contrasting

Unclassified

Order By: Relevance

“…Where LMEs resulted in multiple candidate models, we used a model averaging approach to incorporate model uncertainty and produced model averaged estimates (Burnham & Anderson, 1998, 2002. However, due to the potential pitfalls of using model averaging for ecological data (Banner & Higgs, 2016;Cade, 2015;Harrison et al, 2018), we also present model averaged predictions using candidate models as an additional approach to interpreting the strength and direction of fixed factors (Cade, 2015).…”

Section: Resultsmentioning

confidence: 99%

External temperature and distance from nearest entrance influence microclimates of cave and culvert‐roosting tri‐colored bats (Perimyotis subflavus)

Leivers¹,

Meierhofer

Pierce³

et al. 2019

Ecology and Evolution

View full text Add to dashboard Cite

Many North American bat species hibernate in both natural and artificial roosts. Although hibernacula can have high internal climate stability, they still retain spatial variability in their thermal regimes, resulting in various “microclimates” throughout the roost that differ in their characteristics (e.g., temperature and air moisture). These microclimate components can be influenced by factors such as the number of entrances, the depth of the roost, and distance to the nearest entrance of the roost. Tri‐colored bats are commonly found roosting in caves in winter, but they can also be found roosting in large numbers in culverts, providing the unique opportunity to investigate factors influencing microclimates of bats in both natural and artificial roost sites. As tri‐colored bats are currently under consideration for federal listing, information of this type could be useful in aiding in the conservation and management of this species through a better understanding of what factors affect the microclimate near roosting bats. We collected data on microclimate temperature and microclimate actual water vapor pressure (AWVP) from a total of 760 overwintering tri‐colored bats at 18 caves and 44 culverts. Using linear mixed models analysis, we found that variation in bat microclimate temperatures was best explained by external temperature and distance from nearest entrance in both caves and culverts. External temperature had a greater influence on microclimate temperatures in culverts than caves. We found that variation in microclimate AWVP was best explained by external temperature, distance from nearest entrance, and proportion from entrance (proportion of the total length of the roost from the nearest entrance) in culvert‐roosting bats. Variation in microclimate AWVP was best explained by external temperature and proportion from entrance in cave‐roosting bats. Our results suggest that bat microclimate temperature and AWVP are influenced by similar factors in both artificial and natural roosts, although the relative contribution of these factors differs between roost types.

show abstract

Section: Resultsmentioning

confidence: 99%

External temperature and distance from nearest entrance influence microclimates of cave and culvert‐roosting tri‐colored bats (Perimyotis subflavus)

Leivers¹,

Meierhofer

Pierce³

et al. 2019

Ecology and Evolution

View full text Add to dashboard Cite

show abstract

“…Since different sites were sampled in different years and the number of sampling rounds varied, the dataset was not well suited for a more formal time-series analysis. To control temporal dependency, we used generalised linear mixed models (GLMMs) that included sample site as a random effect, allowing the estimation of variance in the response variables within and among sample sites (Harrison et al, 2018). This allowed us to increase statistical power, when compared to a more formal time-series analysis (Zuur, Ieno, Walker, Saveliev, & Smith, 2009), by modelling a dependency structure among diversity measures through time at the same site.…”

Section: Long-term Changes In Fish Diversity In Lotic and Lentic Enmentioning

confidence: 99%

Temporal changes in fish diversity in lotic and lentic environments along a reservoir cascade

Loures

2019

Freshwater Biology

View full text Add to dashboard Cite

Long‐term modifications in hydrological conditions due to river damming result in varied effects on fish diversity, which can be stronger in a reservoir cascade. Therefore, we assessed changes in fish diversity over a 28‐year period, in a series of reservoirs along the Araguari River Basin, a Neotropical drainage in Brazil, in both lentic environments (reservoirs) and the lotic stretches between them. We used linear mixed models to assess long‐term changes in fish diversity. Diversity was measured as α diversity (fish species richness) and β diversity (Jaccard's and Raup–Crick's dissimilarity index). Beta diversity was then partitioned into turnover (species replacement) and nestedness (species loss). To understand fish species replacement along the cascade, we modelled turnover with four covariates: age, richness of native and non‐native fish species, and longitudinal position of sampling site along the cascade. The reduction of total richness and the increase in non‐native richness were related to reservoir age. These effects were similar in both lotic and lentic environments, with no significant difference between them. We also observed a decrease in native fish richness over time, with a slight tendency for lentic environments to lose more species than lotic environments. Dissimilarity in fish assemblage increased over time and β Raup–Crick results corroborated those derived from β Jaccard, yielding similar significant temporal trends. Changes in fish species composition occurred predominantly due to turnover, compared to nestedness. We observed a greater species loss in reservoirs, suggesting that the transformation from a lotic environment into a lentic had greater impact on fish assemblages than the flow regulation of the lotic stretches downstream of the dams. Our results suggest that species replacement can be explained mainly by the increase of non‐native fish richness, followed by the longitudinal position along the cascade in an upstream–downstream direction, and reservoir age. This study reveals that, even after 2 decades of monitoring, we still observe important changes in fish assemblages, demonstrating the importance of long‐term monitoring to assess biodiversity. Our findings reinforce the importance of long‐term fish monitoring schemes for the investigation of the effects of non‐native fish species on native fish assemblages.

show abstract

“…The random-effects approach has the advantage of assuming region-specific coefficients are not independent, but are instead drawn from a global distribution of selection coefficients with its own mean and variance (Bolker et al, 2009). This assumption can help improve overall parameter estimation, particularly for regions with small amounts of data, through the process of shrinkage (Harrison et al, 2018;Zipkin, DeWan, & Andrew Royle, 2009). There are also machine learning methods such as bagged decision trees and random forest that can be used to account for differences in habitat selection among regions (sensu Fink et al, 2010;Doherty, Evans, Coates, Juliusson, & Fedy, 2016).…”

Section: Discussionmentioning

confidence: 99%

Differential habitat selection in boreal songbirds influences estimates of population size and distribution

Crosby

Bayne

Cumming

et al. 2019

Diversity and Distributions

View full text Add to dashboard Cite

Aim Most large‐scale species distribution models assume spatially constant habitat selection throughout a species' geographic range. However, there is evidence this assumption may not be valid for a number of boreal bird species, which could lead to biased predictions of density and distribution in range‐wide models. Our goal was to test for and quantify differential habitat selection (DHS) in songbirds among regions of the Canadian boreal forest. Location Northern Alberta, western Ontario and southern Quebec, Canada. Methods We used hierarchical analysis of covariance models with region‐specific parameter estimates to test for differential selection of forest attributes among three regions for six boreal bird species. We used the results of these models to quantify intraspecific niche overlap between regions and compared posterior predictive accuracy to models that did not account for DHS. Results We found a generally large standardized effect size (median effect size = 1.674) of region on selection of specific habitat variables for all six species, although there was high variability among species, variables and regional comparisons. The proportion of niche overlap between regions was generally low (mean overlap = 0.309 for all pairwise comparisons), with no spatial pattern to the overlap. Models accounting for DHS had significantly higher posterior predictive accuracy according to the Watanabe–Akaike information criterion. Main Conclusions We found strong evidence for DHS among regions for six boreal songbird species in individual habitat attributes and overall niche space. The higher predictive accuracy of our DHS models suggests that failure to account for spatial variability in habitat selection can lead to biased estimates of density and spatial distribution. Models that did not account for DHS overestimated density relative to DHS models. We conclude that large‐scale species distribution models should account for regional variation in habitat selection in order to obtain accurate estimates of population size and distribution.

show abstract

A brief introduction to mixed effects modelling and multi-model inference in ecology

Cited by 1,472 publications

References 92 publications

External temperature and distance from nearest entrance influence microclimates of cave and culvert‐roosting tri‐colored bats (Perimyotis subflavus)

External temperature and distance from nearest entrance influence microclimates of cave and culvert‐roosting tri‐colored bats (Perimyotis subflavus)

Temporal changes in fish diversity in lotic and lentic environments along a reservoir cascade

Differential habitat selection in boreal songbirds influences estimates of population size and distribution

Contact Info

Product

Resources

About