USGS Science⌕ Search

Geology topics

David R. Anderson

Publications and source records attributed to David R. Anderson.

At least 19 recordsLinked to original sources
USGS2009

Model selection bias and Freedman's paradox

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. ?? The Institute of Statistical Mathematics, Tokyo 2009.

Annals of the Institute of Statistical Mathematics↗
USGS2007

Line transect estimation of population size: the exponential case with grouped data

Gates, Marshall, and Olson (1968) investigated the line transect method of estimating grouse population densities in the case where sighting probabilities are exponential. This work is followed by a simulation study in Gates (1969). A general overview of line transect analysis is presented by Burnham and Anderson (1976). These articles all deal with the ungrouped data case. In the present article, an analysis of line transect data is formulated under the Gates framework of exponential sighting probabilities and in the context of grouped data.

Communications in Statistics - Theory and Methods↗
USGS2007

Concerns regarding a call for pluralism of information theory and hypothesis testing

1. Stephens et al. (2005) argue for 'pluralism' in statistical analysis, combining null hypothesis testing and information-theoretic (I-T) methods. We show that I-T methods are more informative even in single variable problems and we provide an ecological example. 2. I-T methods allow inferences to be made from multiple models simultaneously. We believe multimodel inference is the future of data analysis, which cannot be achieved with null hypothesis-testing approaches. 3. We argue for a stronger emphasis on critical thinking in science in general and less reliance on exploratory data analysis and data dredging. Deriving alternative hypotheses is central to science; deriving a single interesting science hypothesis and then comparing it to a default null hypothesis (e.g. 'no difference') is not an efficient strategy for gaining knowledge. We think this single-hypothesis strategy has been relied upon too often in the past. 4. We clarify misconceptions presented by Stephens et al. (2005) . 5. We think inference should be made about models, directly linked to scientific hypotheses, and their parameters conditioned on data, Prob(Hj| data). I-T methods provide a basis for this inference. Null hypothesis testing merely provides a probability statement about the data conditioned on a null model, Prob(data |H0). 6. Synthesis and applications . I-T methods provide a more informative approach to inference. I-T methods provide a direct measure of evidence for or against hypotheses and a means to consider simultaneously multiple hypotheses as a basis for rigorous inference. Progress in our science can be accelerated if modern methods can be used intelligently; this includes various I-T and Bayesian methods.

Journal of Applied Ecology↗
USGS2007

Some comments on Anderson and Pospahala's correction of bias in line transect sampling

ANDERSON and POSPAHALA (1970) investigated the estimation of wildlife population size using the belt or line transect sampling method and devised a correction for bias, thus leading to an estimator with interesting characteristics. This work was given a uniform mathematical framework in BURNHAM and ANDERSON (1976). In this paper we show that the ANDERSON-POSPAHALA estimator is optimal in the sense of being the (unique) best linear unbiased estimator within the class of estimators which are linear combinations of cell frequencies, provided certain assumptions are met.

Biometrical Journal↗
USGS2006

Comment on 'Are survival rates for northern spotted owls biased?'

Loehle et al. recently estimated survival rates from radio-telemetered northern spotted owls (Strix occidentalis caurina (Merriam, 1898)) and suggested that survival rates estimated for this species from capture-recapture studies were negatively biased, which subsequently resulted in the negatively biased estimates of rates of population change (lambda) reported by Anthony et al. (Wildl. Monogr. No. 163, pp. 1-47 (2006)). We argue that their survival estimates were inappropriate for comparison with capture-recapture estimates because (i) the manner in which they censored radio-telemetered individuals had the potential to positively bias their survival estimates, (ii) their estimates of survival were not valid for evaluating bias, and (iii) the size and distribution of their radiotelemetry study areas were sufficiently different from capture-recapture study areas to preclude comparisons. In addition, their inferences of negative bias in rates of population change estimated by Anthony et al. were incorrect and reflected a misunderstanding about those estimators.

Canadian Journal of Zoology↗
USGS2006

Status and trends in demography of northern spotted owls, 1985-2003

We analyzed demographic data from northern spotted owls (Strix occidentalis caurina) from 14 study areas in Washington, Oregon, and California for 1985-2003. The purpose of our analyses was to provide an assessment of the status and trends of northern spotted owl populations throughout most of their geographic range. The 14 study areas made up approximately 12% of the range of the subspecies and included federal, tribal, private, and mixed federal and private lands. The study areas also included all the major forest types that the subspecies inhabits. The analyses followed rigorous protocols that were developed a priori and were the result of extensive discussions and consensus among the authors. Our primary objectives were to estimate fecundity, apparent survival (??), and annual rate of population change (??) and to determine if there were any temporal trends in these population parameters. In addition to analyses of data from individual study areas, we conducted 2 meta-analyses on each demographic parameter. One meta-analysis was conducted on all 14 areas, and the other was restricted to the 8 areas that constituted the Effectiveness Monitoring Plan for northern spotted owls under the Northwest Forest Plan. The average number of years of reproductive data per study area was 14 (range = 5-19), and the average number of recapture occasions per study area was 13 (range = 4-18). Only 1 study area had <12 years of data. Our results were based on 32,054 captures and resightings of 11,432 banded individuals for estimation of survival and 10,902 instances in which we documented the number of young produced by territorial females. The number of young fledged (NYF) per territorial female was analyzed by testing a suite of a priori models that included (1) effects of age, (2) linear or quadratic time trends, (3) presence of barred owls (Strix varia) in spotted owl territories, and (4) an even-odd year effect. The NYF varied among years on most study areas with a biennial cycle of high reproduction in even-numbered years and low reproduction in odd-numbered years. These cyclic fluctuations did not occur on all study areas, and the even-odd year effect waned during the last 5 years of the study. Fecundity was highest for adults (x??=0.372, SE=0.029), lower for 2-year-olds (x??=0.208, SE=0.032), and very low for 1-year-olds (x??=0.074, SE = 0.029). Fecundity was stable over time for 6 areas (Rainier, Olympic, Warm Springs, H. J. Andrews, Klamath, and Marin), declining for 6 areas (Wenatchee, Cle Elum, Oregon Coast Range, Southern Oregon Cascades, Northwest California, and Simpson), and slightly increasing for 2 areas (Tyee, Hoopa). We found little association between NYF and the proportion of northern spotted owl territories where barred owls were detected, although results were suggestive of a negative effect of barred owls on the Wenatchee and Olympic study areas. The meta-analysis on fecundity indicated substantial annual variability with no increasing or decreasing trends. Fecundity was highest in the mixed-conifer region of eastern Washington (x??=0.560, SE=0.041) and lowest in the Douglas-fir (Pseudotsuga menziesii) region of the Oregon coast (x??=0.306, SE=0.039). We used Cormack-Jolly-Seber open population models and Program MARK to estimate apparent survival rates of owls >1 year old. We found no differences in apparent survival rates between sexes except for 1 area (Marin), which had only 6 years of data. Estimates of apparent survival from individual study areas indicated that there were differences among age classes with adults generally having higher survival than 1- and 2-year-olds. Apparent survival rates ranged from 0.750 (SE=0.026) to 0.886 (SE=0.010) for adults, 0.626 (SE=0.073) to 0.886 (SE=0.010) for 2-year-olds, and 0.415 (SE=0.111) to 0.860 (SE=0.017) for 1-year-olds. These estimates were comparable to survival rates from previous studies on the subspecies. We found evidence for negative time trends in survival rate

Wildlife Monographs↗
USGS2005

Evaluation of trapping-web designs

The trapping web is a method for estimating the density and abundance of animal populations. A Monte Carlo simulation study is performed to explore performance of the trapping web for estimating animal density under a variety of web designs and animal behaviours. The trapping performs well when animals have home ranges, even if the home ranges are large relative to trap spacing. Webs should contain at least 90 traps. Trapping should continue for 5-7 occasions. Movement rates have little impact on density estimates when animals are confined to home ranges. Estimation is poor when animals do not have home ranges and movement rates are rapid. The trapping web is useful for estimating the density of animals that are hard to detect and occur at potentially low densities. ?? CSIRO 2005.

Wildlife Research↗
USGS2003

Small-mammal density estimation: A field comparison of grid-based vs. web-based density estimators

Statistical models for estimating absolute densities of field populations of animals have been widely used over the last century in both scientific studies and wildlife management programs. To date, two general classes of density estimation models have been developed: models that use data sets from capture–recapture or removal sampling techniques (often derived from trapping grids) from which separate estimates of population size ( NÌ‚ ) and effective sampling area ( AÌ‚ ) are used to calculate density ( DÌ‚ = NÌ‚ / AÌ‚ ); and models applicable to sampling regimes using distance-sampling theory (typically transect lines or trapping webs) to estimate detection functions and densities directly from the distance data. However, few studies have evaluated these respective models for accuracy, precision, and bias on known field populations, and no studies have been conducted that compare the two approaches under controlled field conditions. In this study, we evaluated both classes of density estimators on known densities of enclosed rodent populations. Test data sets ( n = 11) were developed using nine rodent species from capture–recapture live-trapping on both trapping grids and trapping webs in four replicate 4.2-ha enclosures on the Sevilleta National Wildlife Refuge in central New Mexico, USA. Additional “saturation” trapping efforts resulted in an enumeration of the rodent populations in each enclosure, allowing the computation of true densities. Density estimates ( DÌ‚ ) were calculated using program CAPTURE for the grid data sets and program DISTANCE for the web data sets, and these results were compared to the known true densities ( D ) to evaluate each model's relative mean square error, accuracy, precision, and bias. In addition, we evaluated a variety of approaches to each data set's analysis by having a group of independent expert analysts calculate their best density estimates without a priori knowledge of the true densities; this “blind” test allowed us to evaluate the influence of expertise and experience in calculating density estimates in comparison to simply using default values in programs CAPTURE and DISTANCE. While the rodent sample sizes were considerably smaller than the recommended minimum for good model results, we found that several models performed well empirically, including the web-based uniform and half-normal models in program DISTANCE, and the grid-based models M b and M bh in program CAPTURE (with AÌ‚ adjusted by species-specific full mean maximum distance moved (MMDM) values). These models produced accurate DÌ‚ values (with 95% confidence intervals that included the true D values) and exhibited acceptable bias but poor precision. However, in linear regression analyses comparing each model's DÌ‚ values to the true D values over the range of observed test densities, only the web-based uniform model exhibited a regression slope near 1.0; all other models showed substantial slope deviations, indicating biased estimates at higher or lower density values. In addition, the grid-based DÌ‚ analyses using full MMDM values for WÌ‚ area adjustments required a number of theoretical assumptions of uncertain validity, and we therefore viewed their empirical successes with caution. Finally, density estimates from the independent analysts were highly variable, but estimates from web-based approaches had smaller mean square errors and better achieved confidence-interval coverage of D than did grid-based approaches. Our results support the contention that web-based approaches for density estimation of small-mammal populations are both theoretically and empirically superior to grid-based approaches, even when sample size is far less than often recommended. In view of the increasing need for standardized environmental measures for comparisons among ecosystems and through time, analytical models based on distance sampling appear to offer accurate density estimation approaches for research studies involving small-mammal abundances.

Ecological Monographs↗
USGS2001

Suggestions for presenting the results of data analyses

We give suggestions for the presentation of research results from frequentist, information-theoretic, and Bayesian analysis paradigms, followed by several general suggestions. The information-theoretic and Bayesian methods offer alternative approaches to data analysis and inference compared to traditionally used methods. Guidance is lacking on the presentation of results under these alternative procedures and on nontesting aspects of classical frequentists methods of statistical analysis. Null hypothesis testing has come under intense criticism. We recommend less reporting of the results of statistical tests of null hypotheses in cases where the null is surely false anyway, or where the null hypothesis is of little interest to science or management.

Journal of Wildlife Management↗
USGS2001

Field trials of line transect methods applied to estimation of desert tortoise abundance

We examine the degree to which field observers can meet the assumptions underlying line transect sampling to monitor populations of desert tortoises ( Gopherus agassizii ). We present the results of 2 field trials using artificial tortoise models in 3 size classes. The trials were conducted on 2 occasions on an area south of Las Vegas, Nevada, where the density of the test population was known. In the first trials, conducted largely by experienced biologists who had been involved in tortoise surveys for many years, the density of adult tortoise models was well estimated (-3.9% bias), while the bias was higher (-20%) for subadult tortoise models. The bias for combined data was -12.0%. The bias was largely attributed to the failure to detect all tortoise models on or near the transect centerline. The second trials were conducted with a group of largely inexperienced student volunteers and used somewhat different searching methods, and the results were similar to the first trials. Estimated combined density of subadult and adult tortoise models had a negative bias (-7.3%), again attributable to failure to detect some models on or near the centerline. Experience in desert tortoise biology, either comparing the first and second trials or in the second trial with 2 experienced biologists versus 16 novices, did not have an apparent effect on the quality of the data or the accuracy of the estimates. Observer training, specific to line transect sampling, and field testing are important components of a reliable survey. Line transect sampling represents a viable method for largescale monitoring of populations of desert tortoise; however, field protocol must be improved to assure the key assumptions are met.

Nevada↗
USGS2001

Kullback-Leibler information in resolving natural resource conflicts when definitive data exist

Conflicts often arise in the management of natural resources. Often they result from differing perceptions, varying interpretations of the law, and self-interests among stakeholder groups (for example, the values and perceptions about spotted owls and forest management differ markedly among environmental groups, government regulatory agencies, and timber industries). We extend the conceptual approach to conflict resolution of Anderson et al. (1999) by using information-theoretic methods to provide quantitative evidence for differing stakeholder positions. Importantly, we assume that relevant empirical data exist that are central to the potential resolution of the conflict. We present a hypothetical example involving an experiment to assess potential effects of a chemical on monthly survival probabilities of the hen clam (Spisula solidissima). The conflict centers on 3 stakeholder positions: 1) no effect, 2) an acute effect, and 3) an acute and chronic effect of the chemical treatment. Such data were given to 18 analytical teams to make independent analyses and provide the relative evidence for each of 3 stakeholder positions in the conflict. The empirical evidence strongly supports only one of the 3 positions in the conflict: the application of the chemical causes acute and chronic effects on monthly survival, following treatment. Formal inference from all the stakeholder positions is provided for the 2 key parameters underlying the hen clam controversy. The estimates of these parameters were essentially unbiased (the relative bias for the control and treatment group's survival probability was -0.857% and 1.400%, respectively) and precise (coefficients of variation were 0.576% and 2.761%, respectively). The advantages of making formal inference from all the models, rather than drawing conclusions from only the estimated best model, is illustrated. Finally, we contrast information-theoretic and Bayesian approaches in terms of how positions in the controversy enter the formal analysis.

Wildlife Society Bulletin↗
USGS1990

Evaluation of line transect sampling based on remotely sensed data from underwater video

We used underwater video in conjunction with the line transect method and a Fourier series estimator to make 13 independent estimates of the density of known populations of bricks lying on the bottom in shallows of Lake Huron. The pooled estimate of density (95.5 bricks per hectare) was close to the true density (89.8 per hectare), and there was no evidence of bias. Confidence intervals for the individual estimates included the true density 85% of the time instead of the nominal 95%. Our results suggest that reliable estimates of the density of objects on a lake bed can be obtained by the use of remote sensing and line transect sampling theory.

Transactions of the American Fisheries Society↗
USGS1985

Problems in estimating age-specific survival rates from recovery data of birds ringed as young

(1) The life table model is frequently employed in the analysis of ringer samples of young in bird populations. The basic model is biologically unrealistic and of little use in making inferences concerning age-specific survival probabilities. (2) This model rests on a number of restrictive assumptions, the failure of which causes serious biases. Several important assumptions are not met with real data and the estimators of age-specific survival are not robust enough to these failures. (3) Five major problems in the use of the life table method are reviewed. Examples are provided to illustrate several of the problems involved in using this method in making inferences about survival rates and its age-specific nature. (4) We conclude that this is an invalid procedure and it should not be used. Furthermore, ringing studies involving only young birds are pointless as regards survival estimation because no valid method exists for estimating age-specific or time-specific survival rates from such data. (5) In our view, inferences about age-specific survival rates are possible only if both young and adult (or young, subadult and adult) age classes are ringed each year for k years ( k ≥ 2).

Journal of Animal Ecology↗
USGS1984

Tests of compensatory vs. additive hypotheses of mortality in mallards

Band recovery data from over 410 000 adult Mallards ( Anas platyrhynchos ) banded in North America between 1950 and 1979 were analyzed to examine the effect of exploitation on annual survival rate. Two extreme hypotheses were defined: completely compensatory, and totally additive, and an explicit mathematical model was presented to represent each hypothesis. Comparison of the values of the log-likelihood functions at their maxima allowed discrimination between the two models (hypotheses). Extensive Monte Carlo studies were made on the performance of the discriminant test and the power of the test under the two extreme hypotheses and a range of intermediate hypotheses. The results of the discriminant tests on the data agreed closely with the expected performance of the test if the true underlying process was compensatory. We rejected the hypothesis of total additivity ( P =.001). Instead, it appears that hunting mortalities are largely compensated for by other forms of mortality.

Ecology↗