David R. Anderson’s research while affiliated with Colorado State University and other places
What is this page?
This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.
Distance sampling is a widely used group of closely related methods for estimating the density and/or abundance of biological populations. The main methods are line-transect sampling and point-transect sampling. In both cases, observer(s) perform a standardized survey along a series of randomly located lines or points, searching for objects of interest (usually animals or clusters of animals). For each object detected, they record the distance from the line or point to the object. Not all objects will be detected, but a fundamental assumption of the basic methods is that all objects that are actually on the line or point are detected. The key to distance sampling analyses is to use the observed distances to fit a detection function that describes how detectability decreases with increasing distance from the transect. The fitted function is used to estimate the average probability of detecting an object; from here, one can readily obtain point and interval estimates for the density and abundance of objects in the survey area. Various extensions to the basic methods allow assumptions to be relaxed, and include methods that integrate capture-recapture and distance sampling, as well as methods to model spatial variation in density.Keywords:line transect;point transect;mark–recapture distance sampling;density surface model;detection probability modeling;sampling survey;wildlife population assessment;density;abundance;population size
We analyzed data of adult mallards (Anas platyrhynchos) banded throughout North America during the past 30 years, using a model developed to estimate heterogeneity in survival rates of marked individuals. The model incorporating heterogeneity in survival rates was often the most parsimonious model for the data sets analyzed, thus constituting an advancement over previous band-recovery models. Likelihood-ratio tests indicated that significant amounts of heterogeneity existed in the data sets analyzed. We were unable to detect significant differences in heterogeneity between sexes or between birds banded prior to the hunting season (July–September) and after the hunting season (January–February) in the same areas. Heterogeneity varied widely among banding areas. Changes in reporting rates confound the estimation of heterogeneity. Adjustment of estimates of heterogeneity for differences in recovery rates in the first year after banding and subsequent years results in lower estimates of heterogeneity for mallards. Coefficients of variation of survival rates for adult mallards, adjusted for variation in recovery rates, varied from 18 to 25%.
We briefly outline the information-theoretic (I-T) approaches to valid inference including a review of some simple methods
for making formal inference from all the hypotheses in the model set (multimodel inference). The I-T approaches can replace
the usual t tests and ANOVA tables that are so inferentially limited, but still commonly used. The I-T methods are easy to compute and
understand and provide formal measures of the strength of evidence for both the null and alternative hypotheses, given the
data. We give an example to highlight the importance of deriving alternative hypotheses and representing these as probability
models. Fifteen technical issues are addressed to clarify various points that have appeared incorrectly in the recent literature.
We offer several remarks regarding the future of empirical science and data analysis under an I-T framework.
KeywordsAIC–Evidence–Kullback–Leibler information–Model averaging–Model likelihoods–Model probabilities–Model selection–Multimodel inference
We used data from 11 long-term studies to assess temporal and spatial patterns in fecundity, apparent survival, recruitment, and annual finite rate of population change of Northern Spotted Owls (Strix occidentalis caurina) from 1985 to 2008. Our objectives were to evaluate the status and trends of the subspecies throughout its range and to investigate associations between population parameters and covariates that might be influencing any observed trends. We examined associations between population parameters and temporal, spatial, and ecological covariates by developing a set of a priori hypotheses and models for each analysis. We used information-theoretic methods and QAIC(c) model selection to choose the best model(s) and rank the rest. Variables included in models were gender, age, and effects of time. Covariates included in some analyses were reproductive success, presence of Barred Owls (Strix varia), percent cover of suitable owl habitat, several weather and climate variables including seasonal and annual variation in precipitation and temperature, and three long-term climate indices. Estimates of fecundity, apparent survival, recruitment, and annual rate of population change were computed from the best models or with model averaging for each study area. The average number of years of reproductive data from each study area was 19 (range = 17 to 24), and the average number of captures/resightings per study area was 2,219 (range = 583 to 3,777), excluding multiple resightings of the same individuals in the same year. The total sample of 5,224 marked owls included 796 1-yr-old subadults, 903 2-yr-old subadults, and 3,545 adults (>= 3 yrs old). The total number of annual captures/recaptures/resightings was 24,408, and the total number of cases in which we determined the number of young produced was 11,450.Age had an important effect on fecundity, with adult females generally having higher fecundity than 1- or 2-yr-old females. Nine of the 11 study areas had an even-odd year effect on fecundity in the best model or a competitive model, with higher fecundity in even years. Based on the best model that included a time trend in fecundity, we concluded that fecundity was declining on five areas, stable on three areas, and increasing on three areas. Evidence for an effect of Barred Owl presence on fecundity on individual study areas was somewhat mixed. The Barred Owl covariate was included in the best model or a competitive model for five study areas, but the relationship was negative for four areas and positive for one area. At the other six study areas, the association between fecundity and the proportion of Spotted Owl territories in which Barred Owls were detected was weak or absent. The percent cover of suitable owl habitat was in the top fecundity model for all study areas in Oregon, and in competitive models for two of the three study areas in Washington. In Oregon, all 95% confidence intervals on beta coefficients for the habitat covariate excluded zero, and on four of the five areas the relationship between the percent cover of suitable owl habitat and fecundity was positive, as predicted. However, contrary to our predictions, fecundity on one of the Oregon study areas (KLA) declined with increases in suitable habitat. On all three study areas in Washington, the beta estimates for the effects of habitat on fecundity had 95% confidence intervals that broadly overlapped zero, suggesting there was less evidence of a habitat effect on fecundity on those study areas. Habitat effects were not included in models for study areas in California, because we did not have a comparable habitat map for those areas. Weather covariates explained some of the variability in fecundity for five study areas, but the best weather covariate and the direction of the effect varied among areas. For example, there was evidence that fecundity was negatively associated with low temperatures and high amounts of precipitation during the early nesting season on three study areas but not on the other eight study areas.The meta-analysis of fecundity for all study areas (no habitat covariates included) suggested that fecundity varied by time and was parallel across ecoregions or latitudinal gradients, with some weak evidence for a negative Barred Owl (BO) effect. However, the 95% confidence interval for the beta coefficient for the BO effect overlapped zero ((beta) over cap = -0.12, SE = 0.11, 95% CI = -0.31 to 0.07). The best models from the meta-analysis of fecundity for Washington and Oregon (habitat covariates included) included the effects of ecoregion and annual time plus weak effects of habitat and Barred Owls. However, the 95% confidence intervals for beta coefficients for the effects of Barred Owls and habitat overlapped zero ((beta) over cap (BO) = -0.104, 95% CI = -0.369 to 0.151; (beta) over cap (HAB1) = -0.469, 95% CI = -1.363 to 0.426). In both meta-analyses of fecundity, linear trends (T) in fecundity were not supported, nor were effects of land ownership, weather, or climate covariates. Average fecundity over all years was similar among ecoregions except for the Washington-Mixed-Conifer ecoregion, where mean fecundity was 1.7 to 2.0 times higher than in the other ecoregions.In the analysis of apparent survival on individual study areas, recapture probabilities typically ranged from 0.70 to 0.90. Survival differed among age groups, with subadults, especially 1-yr-olds, having lower apparent survival than adults. There was strong support for declining adult survival on 10 of 11 study areas, and declines were most evident in Washington and northwest Oregon. There was also evidence that apparent survival was negatively associated with the presence of Barred Owls on six of the study areas. In the analyses of individual study areas, we found little evidence for differences in apparent survival between males and females, or for negative effects of reproduction on survival in the following year.In the meta-analysis of apparent survival, the best model was a random effects model in which survival varied among study areas (g) and years (t), and recapture rates varied among study areas, sexes (s), and years. This model also included the random effects of study area and reproduction (R). The effect of reproduction was negative ((beta) over cap = -0.024), with a 95% confidence interval that barely overlapped zero (-0.049 to 0.001). Several random effects models were competitive, including a second-best model that included the Barred Owl (BO) covariate. The estimated regression coefficient for the BO covariate was negative ((beta) over cap = -0.086), with a 95% confidence interval that did not overlap zero (-0.158 to -0.014). One competitive random effects model included a negative linear time trend on survival ((beta) over cap = -0.0016) with a 95% confidence interval (-0.0035 to 0.0003) that barely overlapped zero. Other random effects models that were competitive with the best model included climate effects (Pacific Decadal Oscillation, Southern Oscillation Index) or weather effects (early nesting season precipitation, early nesting season temperature). Ownership category, percent cover of suitable owl habitat, and latitude had little to no effect on apparent survival. Apparent survival differed among ecoregions, but the ecoregion covariate explained little of the variation among study areas and years.Estimates of the annual finite rate of population change (lambda) were below 1.0 for all study areas, and there was strong evidence that populations on 7 of the 11 study areas declined during the study. For four study areas, the 95% confidence intervals for lambda overlapped 1.0, so we could not conclude that those populations were declining. The weighted mean estimate of lambda for all study areas was 0.971 (SE = 0.007, 95% CI = 0.960 to 0.983), indicating that the average rate of population decline in all study areas combined was 2.9% per year. Annual rates of decline were most precipitous on study areas in Washington and northern Oregon. Based on estimates of realized population change, populations on four study areas declined 40 to 60% during the study, and populations on three study areas declined 20 to 30%. Declines on the other four areas were less dramatic (5 to 15%), with 95% confidence intervals that broadly overlapped 1.0.Based on the top-ranked a priori model in the meta-analysis of lambda, there was evidence that ecoregions and the proportion of Spotted Owl territories with Barred Owl detections were important sources of variation for apparent survival (phi(t)) and recruitment (f(t)). There was some evidence that recruitment was higher on study areas dominated by federal lands compared to study areas that were on private lands or lands that included approximately equal amounts of federal and private lands. There also was evidence that recruitment was positively related to the proportion of the study area that was covered by suitable owl habitat.We concluded that fecundity, apparent survival, and/or populations were declining on most study areas, and that increasing numbers of Barred Owls and loss of habitat were partly responsible for these declines. However, fecundity and survival showed considerable annual variation at all study areas, little of which was explained by the covariates that we used. Although our study areas were not randomly selected, we believe our results reflected conditions on federal lands and areas of mixed federal and private lands within the range of the Northern Spotted Owl because the study areas were (1) large, covering approximate to 9% of the range of the subspecies; (2) distributed across a broad geographic region and within most of the geographic provinces occupied by the owl; and (3) the percent cover of owl habitat was similar between our study areas and the surrounding landscapes.
Abstract Although several studies have indicated the importance of forbs in brood habitats, no studies have quantified direct effects of the amount of forb cover on sage-grouse (Centrocercus urophasianus) chicks. In 2002 and 2003, we conducted field experiments in Middle Park and Moffat County, Colorado, USA, respectively. Our objective was to quantify effects of 3 levels of forb cover in brood habitat on mass gain and feather growth of human-imprinted sage-grouse chicks. The results indicate that increasing forb cover in brood areas with <20% forb cover may lead to increased chick survival and grouse productivity.
A general, consistent strategy for data analysis is outlined, based on information and likelihood theory. A priori considerations lead to the definition of a set of candidate models, simple criteria are useful in ranking and calibrating the models based on estimates of (relative) Kullback-Leibler information, inference can be based on either the best model or a weighted average of several models. Model selection uncertainty can be quantified and should be incorporated into estimators of precision. Some comments are offered on statistical hypothesis testing and data dredging.
In situations where limited knowledge of a system exists and the ratio of data points to variables is small, variable selection
methods can often be misleading. Freedman (Am Stat 37:152–155, 1983) demonstrated how common it is to select completely unrelated
variables as highly “significant” when the number of data points is similar in magnitude to the number of variables. A new
type of model averaging estimator based on model selection with Akaike’s AIC is used with linear regression to investigate
the problems of likely inclusion of spurious effects and model selection bias, the bias introduced while using the data to
select a single seemingly “best” model from a (often large) set of models employing many predictor variables. The new model
averaging estimator helps reduce these problems and provides confidence interval coverage at the nominal level while traditional
stepwise selection has poor inferential properties.
Counting techniques are widely used to study and monitor terrestrial birds. To assess current applications of counting techniques, we reviewed landbird studies published 1989-1998 in nine major journals and one symposium. Commonly used techniques fell into two groups: Procedures that used counts of bird detections as an index to abundance (index counts), and procedures that used empirical models of detectability to estimate density. Index counts rely upon assumptions concerning detectability that are difficult or impossible to meet in most field studies, but nonetheless remain the technique of choice among ornithologists; 95% of studies we reviewed relied upon point counts, strip transects, or other index procedures. Detectability-based density estimates were rarely used and deserve wider application in landbird studies. Distance sampling is a comprehensive extension of earlier detectability-based procedures (variable-width transects, variable circular plots) and a viable alternative to index counts. We provide a conceptual overview of distance sampling, specific recommendations for applying this technique to studies of landbirds, and an introduction to analysis of distance sampling data using the program DISTANCE.
The development of statistical methods for the analysis of demographic processes in marked animal populations has brought
with it the challenges of communication between the disciplines of statistics, ecology, evolutionary biology and computer
science. In order to aid communication and comprehension, we sought to root out a number of cases of ambiguity, redundancy
and inaccuracy in notation and terminology that have developed in the literature. We invited all working in this field to
submit topics for resolution and to express their own views. In the ensuing discussion forum it was then possible to establish
a series of general principles which were, almost without exception, unanimously accepted. Here we set out the background
to the areas of confusion, how these were debated and the conclusions which were reached in each case. We hope that the resulting
guidelines will be widely adopted as standard terminology in publications and in software for the analysis of demographic
processes in marked animal populations
KeywordsMark-recapture–Mark-recovery–Terminology–Notation
Science is about discovering new things, about better understanding processes and systems, and generally furthering our knowledge. Deep in science philosophy is the notion of hypotheses and mathematical models to represent these hypotheses. It is partially the quantification of hypotheses that provides the illusive concept of rigor in science. Science is partially an adversarial process; hypotheses battle for primacy aided by observations, data, and models. Science is one of the few human endeavors that is truly progressive. Progress in science is defined as approaching an increased understanding of truth – science evolves in a sense.
Citations (80)
... On the other hand, as we hypothesized, moving territory as well as the distance moved are mainly associated with the loss of the partner in the previous year. This finding is in line with the idea that species with lasting pair bonding and marked territorial behavior are reluctant to shift to another breeding territory unless a mate dies (Blakesley et al. 2006;Seamans and Gutiérrez 2007;Jenkins et al. 2019). In tawny owls, both sexes are territorial all year round and involved in territory defense (Appleby et al. 1999;Sunde 2011). ...
... Avian responsiveness to environmental change (Furness and Greenwood 1993), the various ecosystem services they provide (Whelan et al. 2015), and the feasibility of bird surveys relative to other taxa (Bibby et al. 2000) have made counting birds attractive for investigating both population-level and macroecological questions (Wiens and Rotenberry 1985, Brown 1995, Hanski 1998. Historically, ornithologists primarily counted birds following standardized sampling designs (Table 1) and survey protocols (Table 1; Ralph et al. 1993, Rosenstock et al. 2002, Matsuoka et al. 2014) and ...
... Because our procedure results in an estimate for each day of our survey, we took the mean value from May 21 through the end of our survey on May 31. We used the coefficient of variation from our previous bootstrap procedure to construct log-based confidence intervals (Burnham et al. 1987;Buckland et al. 2001) for our estimates. We used the delta method (Dorfman 1938) to calculate standard errors for abundance estimates associated with Bengtson et al. (2005) surveys. ...
... For the High CP group both statistical fit indices and theoretical/clinical usefulness were also considered to select the best model. Although the LMR test did not indicate a significant improvement with the addition of a third profile, the entropy value (0.76) was similar, and the AIC and BIC indices were lower, suggesting a better balance between fit and complexity [91]. For this group, literature also supports the three-profile model, as it offers a more detailed classification consistent with the complexity of CP-related variables, whilst adding one more profile did not favor interpretability. ...
... Scott's algorithm (Scott, 2002) determined the expected number of individuals in each state, which was then compared with observed values and iterated multiple times until convergence was achieved. Variability may be underestimated using this method (Burnham & Anderson, 1998), therefore SEs and CIs were derived by parametric bootstrap (Appendix S2). This method tested variability in replicate fits by simulating and comparing one hundred replicate datasets. ...
... The AIC includes a penalty for over-fitting the model, not allowing for an increase in the statistical bias when more parameters are fitted (Wilson et al. 2013). Another advantage of the AIC in model selection is that AIC is independent of the order in which models are computed (Anderson et al. 2001). ...
... We included 'condition' (blue-enriched or blue-depleted light), 'sex' and the interaction between 'condition' and 'sex' (condition*sex) as predictor variables in a possible set of models (see Table 2 for model comparison). The best overall model was assessed using the Akaike Information Criterion corrected for small sample sizes (AICc) 26 . A Wald t-distribution approximation method was used to calculate 95% confidence intervals and p-values. ...
... There was not significant multicollinearity between covariates, nor overdispersion found for the full model (overdispersion ratio = 0.97). Using an Akaike Information Criterion (AIC)-based model comparison approach, models with varying combinations of explanatory variables and interaction effects were evaluated (Burnham and Anderson 2002). The simplest model, which included only main effects (R code equation: NLP presence ~ health + Sex + Mode Age + (1 | Animal ID)) had the lowest AIC value (∆AIC of 2.99 compared to the full model) and was chosen for analysis (Burnham and Anderson 2002). ...
... All possible subsets model selection and model averaging was then carried out using the 'dredge' function from the MuMIn package (Bartoń 2023). Selection was done based on model AICc and models were considered equally parsimonious when within ΔAICc < 2 (Burnham and Anderson 2002). The adjusted criterion (AICc) was used due to a relatively low sample size (N = 159) resulting in a data to estimated model parameter ratio being > 40 (Burnham and Anderson 2002). ...
... R 2 and RSE) and that of predictor power (AICc, ∆AICc, and AICcWt) and potential differences that occurred between the two methods [21]. Burnham and Anderson [18] demonstrated the benefits of using AIC as a one-dimensional method of model selection with a standard set of criteria and the need to optimize the use of biologically relevant predictor variables. Furthermore, the typical use of multiple regressions with many parameters should be considered the global model (the "inclusive" model) and not necessarily the optimal model, while more simplistic models may demonstrate greater biological relevance. ...