Performance tradeoffs in target‐group bias correction for species distribution models

Species distribution models (SDMs) are often calibrated using presence‐only datasets plagued with environmental sampling bias, which leads to a decrease of model accuracy. In order to compensate for this bias, it has been suggested that background data (or pseudoabsences) should represent the area that has been sampled. However, spatially‐explicit knowledge of sampling effort is rarely available. In multi‐species studies, sampling effort has been inferred following the target‐group (TG) approach, where aggregated occurrence of TG species informs the selection of background data. However, little is known about the species‐ specific response to this type of bias correction. The present study aims at evaluating the impacts of sampling bias and bias correction on SDM performance. To this end, we designed a realistic system of sampling bias and virtual species based on 92 terrestrial mammal species occurring in the Mediterranean basin. We manipulated presence and background data selection to calibrate four SDM types. Unbiased (unbiased presence data) and biased (biased presence data) SDMs were calibrated using randomly distributed background data. We used real and TG‐estimated sampling efforts in background selection to correct for sampling bias in presence data. Overall, environmental sampling bias had a deleterious effect on SDM performance. In addition, bias correction improved model accuracy, and especially when based on spatially‐explicit knowledge of sampling effort. However, our results highlight important species‐specific variations in susceptibility to sampling bias, which were largely explained by range size: widely‐distributed species were most vulnerable to sampling bias and bias correction was even detrimental for narrow‐ranging species. Furthermore, spatial discrepancies in SDM predictions suggest that bias correction effectively replaces an underestimation bias with an overestimation bias, particularly in areas of low sampling intensity. Thus, our results call for a better estimation of sampling effort in multispecies system, and cautions the uninformed and automatic application of TG bias correction.     

Understanding reporting bias in the dietary recall data of 11-year-old girls

Objective: This study describes patterns of bias in self‐reported dietary recall data of girls by examining differences among girls classified as under‐reporters, plausible reporters, and over‐reporters on weight, dietary patterns, and psychosocial characteristics. Research Methods and Procedures: Participants included 176 girls at age 11 and their parents. Girls’ weight and height were measured. Three 24‐hour dietary recalls and responses to psychosocial measures were collected. Plausibility cut‐offs for reported energy intake as a percentage of predicted energy requirements were used to divide the sample into under‐reporters, plausible reporters, and over‐reporters. Differences among these three groups on dietary and psychosocial variables were assessed to examine possible sources of bias in reporting. Results: Using a ±1 standard deviation cut‐off for energy intake plausibility, 50% of the sample was categorized as plausible reporters, 34% as under‐reporters, and 16% as over‐reporters. Weight status of under‐reporters was significantly higher than that of plausible reporters and over‐reporters. With respect to reported dietary intake, under‐reporters were no different from plausible reporters on intakes of foods with higher nutrient densities and lower energy densities and were significantly lower than plausible reporters on intakes of foods with lower nutrient densities and higher energy densities. Over‐reporters reported significantly higher intakes of all food groups and the majority of subgroups, relative to plausible reporters. Under‐reporters had significantly higher levels of weight concern and dietary restraint than both plausible reporters and over‐reporters. Discussion: Techniques to categorize plausible and implausible reporters can and should be used to provide an improved understanding of the nature of error in children's dietary intake data and account for this error in analysis and interpretation.     

Inferences from quantitative analysis of dental microwear

Tooth microwear studies have been carried out for several reasons in the last decade. Most effort has been put into categorizing wear types that reflect dietary preferences in order to reconstruct the diet of extinct species. Several studies have shown that, for primates, carnivores and ruminants, it is possible to differentiate statistically the microwear associated with the major dietary adaptations in the group. It has further been found that more subtle dietary changes, such as seasonal ones, can be picked up if the sampling is good enough. It is important to recognize that, although it may be a valuable and legitimate concern to study the specific causes of different microwear patterns, that information is not essential for dietary reconstruction, if different microwear states can be shown empirically to correspond to different dietary regimes. Image enhancement and optical diffraction methods offer hope of automated scanning of large samples. This will be a major benefit as the methods currently in use are labor-intensive and time-consuming. Finally, it is urged that as many methods as possible be used to solve problems of dietary reconstruction.     

Phylogenetic utility of elongation factor-1 alpha in noctuoidea (Insecta: Lepidoptera): the limits of synonymous substitution

To test its phylogenetic utility, nucleotide sequence variation in a 1,240-bp fragment of the elongation factor-1 alpha (EF-1 alpha) gene was examined in 49 moth species representing the major groups of the superfamily Noctuoidea. Both parsimony and distance analyses supported the monophyly of nearly all groups for which there are clear morphological synapomorphies. Clades of subfamily rank and lower, probably mid-Tertiary and younger, were strongly supported. The third codon position contains 88% of variable sites, and approaches saturation at approximately 20% sequence divergence, possibly due to among-site rate heterogeneity and composition bias; higher divergences occur only in association with shifts in composition. Surprisingly, the few nonsynonymous changes appear no more phylogenetically reliable than synonymous changes. Signal strength for basal divergences is weak and fails to improve with character weighting; thus, dense taxon sampling is probably needed for strong inference from EF-1 alpha regarding deeper splits in Noctuoidea (probably early Tertiary). EF-1 alpha synonymous changes show promise for phylogeny reconstruction within Noctuidae and other groups of Tertiary age.      

Below-threshold mortality: implications for studies in evolution,ecology and demography

Evolutionary biologists, ecologists and experimental gerontologists have increasingly used estimates of age-specific mortality as a critical component in studies of a range of important biological processes. However, the analysis of age-specific mortality rates is plagued by specific statistical challenges caused by sampling error. Here we discuss the nature of this ‘demographic sampling error’, and the way in which it can bias our estimates of (1) rates of ageing, (2) age at onset of senescence, (3) costs of reproduction and (4) demographic tests of evolutionary models of ageing. We conducted simulations which suggest that using standard statistical techniques, we would need sample sizes on the order of tens of thousands in most experiments to effectively remove any bias due to sampling error. We argue that biologists should use much larger sample sizes than have previously been used. However, we also present simple maximum likelihood models that effectively remove biases due to demographic sampling error even at relatively small sample sizes.     

Improved species‐occurrence predictions in data‐poor regions: using large‐scale data and bias correction with down‐weighted Poisson regression and Maxent

Species distribution modelling (SDM) has become an essential method in ecology and conservation. In the absence of survey data, the majority of SDMs are calibrated with opportunistic presence‐only data, incurring substantial sampling bias. We address the challenge of correcting for sampling bias in the data‐sparse situations. We modelled the relative intensity of bat records in their entire range using three modelling algorithms under the point‐process modelling framework (GLMs with subset selection, GLMs fitted with an elastic‐net penalty, and Maxent). To correct for sampling bias, we applied model‐based bias correction by incorporating spatial information on site accessibility or sampling efforts. We evaluated the effect of bias correction on the models’ predictive performance (AUC and TSS), calculated on spatial‐block cross‐validation and a holdout data set. When evaluated with independent, but also sampling‐biased test data, correction for sampling bias led to improved predictions. The predictive performance of the three modelling algorithms was very similar. Elastic‐net models have intermediate performance, with slight advantage for GLMs on cross‐validation and Maxent on hold‐out evaluation. Model‐based bias correction is very useful in data‐sparse situations, where detailed data are not available to apply other bias correction methods. However, bias correction success depends on how well the selected bias variables describe the sources of bias. In this study, accessibility covariates described bias in our data better than the effort covariate, and their use led to larger changes in predictive performance. Objectively evaluating bias correction requires bias‐free presence–absence test data, and without them the real improvement for describing a species’ environmental niche cannot be assessed.     

Life table assay of field-caught Mediterranean fruit flies, Ceratitis capitata, reveals age bias

Though traps are used widely to sample phytophagous insects for research or management purposes, and recently in aging research, possible bias stemming from differential response of individuals of various ages to traps has never been examined. In this paper, we tested the response of Ceratitis capitata (Wiedemann) (Diptera: Tephritidae) males and females of four ages (spanning from 1 to 40 days) to McPhail-type traps baited with a synthetic food attractant in field cages and found that the probability of trapping was significantly influenced by age. The type of food on which flies were maintained before testing (sugar or protein) also had a strong effect and interacted with age. In another experiment, we collected wild C. capitata adults of unknown age using 1–3 methods and then reared them in the laboratory until death. The survival schedules of these flies were subsequently used in a life table assay to infer their age at the time of capture. Results showed that on a single sampling date, males captured in traps baited with a food attractant were younger compared with males aspirated from fruiting host trees, or males captured in traps baited with a sex attractant. Likewise, females captured in food-baited traps were younger compared with aspirated females. In addition to providing the first evidence of age-dependent sampling bias for a phytophagous insect species, this paper also provides a novel approach to estimate the differences in the age composition of samples collected with different techniques. These findings are of utmost importance for several categories of insects, medically important groups notwithstanding.     

Sampling bias and model choice in continuous phylogeography: Getting lost on a random walk

Phylogeographic inference allows reconstruction of past geographical spread of pathogens or living organisms by integrating genetic and geographic data. A popular model in continuous phylogeography—with location data provided in the form of latitude and longitude coordinates—describes spread as a Brownian motion (Brownian Motion Phylogeography, BMP) in continuous space and time, akin to similar models of continuous trait evolution. Here, we show that reconstructions using this model can be strongly affected by sampling biases, such as the lack of sampling from certain areas. As an attempt to reduce the effects of sampling bias on BMP, we consider the addition of sequence-free samples from under-sampled areas. While this approach alleviates the effects of sampling bias, in most scenarios this will not be a viable option due to the need for prior knowledge of an outbreak’s spatial distribution. We therefore consider an alternative model, the spatial Λ-Fleming-Viot process (ΛFV), which has recently gained popularity in population genetics. Despite the ΛFV’s robustness to sampling biases, we find that the different assumptions of the ΛFV and BMP models result in different applicabilities, with the ΛFV being more appropriate for scenarios of endemic spread, and BMP being more appropriate for recent outbreaks or colonizations.     

Accuracy of weighed dietary records in studies of diet and health.

OBJECTIVE--To provide an independent evaluation of seven day weighed dietary records, which are currently accepted as the most accurate technique for assessing habitual dietary intake in studies investigating the links between diet and health. DESIGN--Subjects who had previously participated in the Northern Ireland diet and health study were reselected by stratified random sampling to represent the range of energy intakes in the study as assessed by the seven day weighed dietary record. SETTING--Northern Ireland. SUBJECTS--31 Free living adults (16 men and 15 women). MAIN OUTCOME MEASURES--Energy intake as measured by the seven day weighed dietary record and total energy expenditure estimated concurrently by the doubly labelled water technique. RESULTS--Average recorded energy intakes were significantly lower than measured expenditure in the group overall (9.66 MJ/day v 12.15 MJ/day, 95% confidence interval 1.45 to 3.53 MJ/day). Among those in the upper third of energy intakes the mean (SE) ratio of intake to expenditure was close to 1.0, indicating accurate records (men 1.01 (0.11), women 0.96 (0.08]. In the middle and lower thirds the ratios for men were only 0.74 (0.05) and 0.70 (0.07) respectively and for women 0.89 (0.07) and 0.61 (0.07). CONCLUSIONS--These results show a serious bias in reporting habitual energy intake. If substantiated they may have wide implications for the interpretation of many nutritional studies.     

Spatial association of genetically similar Atlantic salmon juveniles and sex bias in spatial patterns in a river

Atlantic salmon ( Salmo salar ) juveniles were electrofished along a 300-m stretch of a river to test for possible associations between genetic similarity and geographical distance between individuals. Multilocus DNA fingerprinting showed that genetically similar juveniles (1–4 years old) were found closer together in the river than less related individuals. However, the association between genetic similarity and geographical distance, although significant, was not strong. This may indicate that factors other than genetic relatedness influence the positioning in the river. A sex bias in the relationship between genetic similarity and geographical distance was caused by a difference between sexually mature and immature males. The study shows that sampling of salmon juveniles should be spread over a wide stretch of the river in order to avoid sampling relatives. Moreover, by including several year classes, the overall degree of genetic similarity is effectively reduced compared with sampling individuals of similar age.     

Detection, identification, and correction of a bias in an epidemiological study

The relative lack of epidemiological studies of natural populations is partly due to the difficulty of obtaining samples that are both large enough and representative of the population. Here, we present the result of an epidemiological study (December 1992-August 1995) of feline immunodeficiency virus (FIV) in a free-roaming population of domestic cats (Felis catus), with a special emphasis on sample bias. Over five trapping periods, the prevalence of FIV in sampled cats steadily declined. Across these samples we consistently achieved a very large sampling fraction (approximately 60% of the population), the sex ratio, age and weight distributions remained stable with time in the samples, and the sex ratio was similar in the samples and the population. These indices would normally indicate that our samples were representative, suggesting the decline in FIV prevalence to be real. However, a concomitant ecological study of the whole population revealed an important bias in the samples, with an initial high probability of capturing a few individuals, which appeared significantly more likely to be FIV-infected, and then a lower probability of recapturing them. Since our protocol resulted in a non-random sampling, subsequent trappings were designed to avoid this bias, by also capturing individuals who had previously learned to escape capture. This modified capture regime revealed that FIV prevalence was in fact constant in the population. This study shows how samples of large size, which are stable and appear representative of the population, can still be biased. These results may have major implications for other studies based on trapping.     

Statistical issues in a metaregression analysis of randomized trials: impact on the dietary sodium intake and blood pressure relationship

In a meta-analysis of randomized trials of the effects of dietary sodium interventions on blood pressure, we found substantial heterogeneity among the studies. We were interested in evaluating whether measurement error, known to be a problem for dietary sodium measures, publication bias, or confounding factors could be responsible for the heterogeneity. A measurement error correction was developed that corrects both the slope and the intercept and takes into account the sample size of each study and the number of measurements taken on an individual. The measurement error correction had a minimal effect on the estimates, although it performed well in simulated data. A smoothed scatter plot was used to assess publication bias. Metaregressions provide a convenient way to jointly assess the effects of several factors, but care must be taken to fit an appropriate model.     

Influence of tail versus cardiac sampling on blood glucose and lipid profiles in mice

Blood is collected during animal experimentation to measure haematological and metabolic parameters. It cannot be assumed that circulating blood has the same composition irrespective of its location, and indeed, differences in the composition of blood sampled from the arterial and venous compartments have been reported. Here we investigated whether blood collected by cardiac puncture (CP) versus that collected following removal of the distal 1 mm of the tail tip (TT) differs with respect to glucose and lipid profiles in male C57BL/6J mice at 4, 7, 20 and 28 weeks of age. Blood was first collected from the TT of unanaesthetized mice, which were then immediately anaesthetized using ketamine/xylazine, and a second blood sample was collected by CP. The CP glucose concentration was significantly higher than TT glucose by a positive bias averaging +80% (P < 0.01), irrespective of the age of the mice. Conversely, the concentrations of the CP lipids, including total cholesterol, high-density lipoprotein cholesterol and triglyceride were lower than TT lipids by a negative bias averaging -25% (P < 0.05). These observations highlight the difficulty in measuring and comparing metabolic parameters such as glucose and lipid between one blood compartment and another. They illustrate the need to standardize sampling sites, especially when repeated blood sampling is required.     

Temporal estimates of effective population size in species with overlapping generations

The standard temporal method for estimating effective population size (N(e)) assumes that generations are discrete, but it is routinely applied to species with overlapping generations. We evaluated bias in the estimates N(e) caused by violation of this assumption, using simulated data for three model species: humans (type I survival), sparrow (type II), and barnacle (type III). We verify a previous proposal by Felsenstein that weighting individuals by reproductive value is the correct way to calculate parametric population allele frequencies, in which case the rate of change in age-structured populations conforms to that predicted by discrete-generation models. When the standard temporal method is applied to age-structured species, typical sampling regimes (sampling only newborns or adults; randomly sampling the entire population) do not yield properly weighted allele frequencies and result in biased N(e). The direction and magnitude of the bias are shown to depend on the sampling method and the species' life history. Results for populations that grow (or decline) at a constant rate paralleled those for populations of constant size. If sufficient demographic data are available and certain sampling restrictions are met, the Jorde-Ryman modification of the temporal method can be applied to any species with overlapping generations. Alternatively, spacing the temporal samples many generations apart maximizes the drift signal compared to sampling biases associated with age structure.     

Dietary breadth of frugivorous birds in relation to their feeding strategies in the lowland forests of central Honshu,Japan

Variation in the dietary breadth of frugivorous birds, i.e. variation in the degree of dietary specialization, is a key factor for understanding the construction of the architecture of diffused fruit–bird interactions. However, little information is available on the dietary breadth of frugivorous birds in a community and the avian attributes influencing it. In this study, we evaluated and compared the dietary breadth of 23 frugivorous birds (including both seed dispersers and seed predators) in relation to their feeding strategies (depending on which the birds were categorized into four types as gulpers, grinders, crushers and peckers) and other avian attributes. In particular, we hypothesized that feeding types with long fruit‐handling time (i.e. crushers and peckers) have a narrow dietary breadth. Our analysis was based on the data obtained through long‐term volunteer monitoring in the lowland forests of Kanagawa prefecture, Japan. We evaluated the dietary breadth of the birds while controlling for differences in sampling efforts and seasonal variation in the available fruit assemblages among the birds. Our analysis revealed that there was a large variation in dietary breadth among the birds, even after controlling for differences in sampling efforts and seasonal variation in the available fruit assemblages. Feeding strategies of the birds defined their dietary breadth to some degree, but we found little support for the hypothesis that fruit handling determines dietary breadth. We also observed a large variation in dietary breadth among the gulpers, which act mainly as seed dispersers for plants, and this variation could not be explained fully by other avian attributes. Our results showed that variation in the dietary breadth of frugivorous birds is partly determined by their feeding strategies and suggested that other unknown factors may play a role in determining dietary breadth variation and in structuring fruit–bird interaction networks.     

The relationship between zinc and copper status and lipid levels in African-Americans

Studies examining the role of zinc and copper nutriture as risk factors for cardiovascular disease in European Americans have produced conflicting results. This study assessed the associations between zinc and copper status and serum lipid levels in an adult African-American community. A cross-sectional study was conducted on 600 individuals (233 males, 367 females) from 25 to 65 yr of age using a random sampling design in a small city in Alabama. Anthropometric, dietary, and serum zinc, copper, and lipid measurements were made. The mean serum zinc and copper levels and dietary zinc intake were similar to that reported previously for European Americans. There were no significant associations between serum zinc, copper, or zinc/copper ratio and total serum cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), or triglyceride levels. For males, there was a small but significant association between dietary zinc and the total cholesterol/HDL-C ratio (r=−0.17, p=0.03). Similarly, females taking either zinc supplements or a multivitamin including zinc had higher HDL-C values than nonsupplementing females. Further prospective studies of the relationship between zinc status and lipid levels in African Americans are needed to verify these results.     

The Effect of Animal Movement on Line Transect Estimates of Abundance

Line transect sampling is a distance sampling method for estimating the abundance of wild animal populations. One key assumption of this method is that all animals are detected at their initial location. Animal movement independent of the transect and observer can thus cause substantial bias. We present an analytic expression for this bias when detection within the transect is certain (strip transect sampling) and use simulation to quantify bias when detection falls off with distance from the line (line transect sampling). We also explore the non-linear relationship between bias, detection, and animal movement by varying detectability and movement type. We consider animals that move in randomly orientated straight lines, which provides an upper bound on bias, and animals that are constrained to a home range of random radius. We find that bias is reduced when animal movement is constrained, and bias is considerably smaller in line transect sampling than strip transect sampling provided that mean animal speed is less than observer speed. By contrast, when mean animal speed exceeds observer speed the bias in line transect sampling becomes comparable with, and may exceed, that of strip transect sampling. Bias from independent animal movement is reduced by the observer searching further perpendicular to the transect, searching a shorter distance ahead and by ignoring animals that may overtake the observer from behind. However, when animals move in response to the observer, the standard practice of searching further ahead should continue as the bias from responsive movement is often greater than that from independent movement.     

Tracking dietary transitions in weanling baboons (Papio hamadryas anubis) using strontium/calcium ratios in enamel

Strontium and calcium are incorporated into developing teeth in a manner that reflects changing physiological concentrations in the body. A new model predicts changes in strontium/calcium (Sr/Ca) ratios in response to dietary transitions experienced at birth and during the weaning period. Microsampling of longitudinal thin sections of tooth enamel using laser ablation inductively coupled plasma mass spectrometry provides a basis for the systematic evaluation of variation in Sr/Ca ratios within the tooth crown. Incremental growth markers in enamel are used to determine the age of onset of enamel mineralization at each sampling point. Thin sections of 5 teeth from 2 wild-caught baboons (Papio hamadryas anubis) were systematically analysed using this technique. Intra- and intertooth analyses of Sr/Ca ratios reveal a pattern of dietary development during the period of enamel formation that is consistent with observational data on the timing of weaning behaviour in anubis baboons.     

Robustness of close‐kin mark–recapture estimators to dispersal limitation and spatially varying sampling probabilities

1. Close‐kin mark–recapture (CKMR) is a method for estimating abundance and vital rates from kinship relationships observed in genetic samples. CKMR inference only requires animals to be sampled once (e.g., lethally), potentially widening the scope of population‐level inference relative to traditional monitoring programs.
2. One assumption of CKMR is that, conditional on individual covariates like age, all animals have an equal probability of being sampled. However, if genetic data are collected opportunistically (e.g., via hunters or fishers), there is potential for spatial variation in sampling probability that can bias CKMR estimators, particularly when genetically related individuals stay in close proximity.
3. We used individual‐based simulation to investigate consequences of dispersal limitation and spatially biased sampling on performance of naive (nonspatial) CKMR estimators of abundance, fecundity, and adult survival. Population dynamics approximated that of a long‐lived mammal species subject to lethal sampling.
4. Naive CKMR abundance estimators were relatively unbiased when dispersal was unconstrained (i.e., complete mixing) or when sampling was random or subject to moderate levels of spatial variation. When dispersal was limited, extreme variation in spatial sampling probabilities negatively biased abundance estimates. Reproductive schedules and survival were well estimated, except for survival when adults could emigrate out of the sampled area. Incomplete mixing was readily detected using Kolmogorov–Smirnov tests.
5. Although CKMR appears promising for estimating abundance and vital rates with opportunistically collected genetic data, care is needed when dispersal limitation is coupled with spatially biased sampling. Fortunately, incomplete mixing is easily detected with adequate sample sizes. In principle, it is possible to devise and fit spatially explicit CKMR models to avoid bias under dispersal limitation, but development of such models necessitates additional complexity (and possibly additional data). We suggest using simulation studies to examine potential bias and precision of proposed modeling approaches prior to implementing a CKMR program.

     

Surgical options for the early-stage breast cancer: factors associated with patient choice and postoperative quality of life.

Patients with early-stage breast cancer have three surgical options: lumpectomy with radiotherapy, mastectomy alone, and mastectomy with breast reconstruction. Our objective was to compare women in these three groups with respect to demographics, preoperative counseling, postoperative body image, and quality of life. Women having undergone surgery for stage 1 or 2 breast cancer between 1990 and 1995 were selected by random sampling of hospital tumor registries and were mailed a self-administered questionnaire, which included the Medical Outcomes Survey Short Form 36. Patients were stratified into three mutually exclusive groups: lumpectomy with axillary node dissection and radiotherapy, modified radical mastectomy, and modified radical mastectomy with breast reconstruction. In total, 267 of 525 surveys were returned (50.9 percent). Compared with mastectomy patients, breast reconstruction patients were younger (p < 0.001), better educated (p = 0.001), and more likely Caucasian (p = 0.02). Among mastectomy patients, 54.9 percent recalled that lumpectomy had been discussed preoperatively and 39.7 percent recalled discussion of breast reconstruction. Post-operative comfort with appearance was significantly lower for mastectomy patients. The relationship between type of surgery and postoperative quality of life varied with age. Under 55, quality of life was lowest for mastectomy patients on all but two Medical Outcomes Survey Short Form 36 subscales. Over 55, quality of life was lowest for lumpectomy patients on all subscales (p < 0.05 for all subscales except social functioning and role-emotional). Treatment choice may be related to age, race, education, and preoperative counseling. Whereas the effect of breast cancer on a woman's life is complex and individual, the type of surgery performed is a significant variable, whose impact may be related to patient age.