Effect Of Sampling Design Research Articles

Toward a New Approach to Creating Population-Representative Data for Demographic Research.

The evaluation of innovative web-based data collection methods that are convenient for the general public and that yield high-quality scientific information for demographic researchers has become critical. Web-based methods are crucial for researchers with nationally representative research objectives but without the resources of larger organizations. The web mode is appealing because it is inexpensive relative to in-person and telephone modes, and it affords a high level of privacy. We evaluate a sequential mixed-mode web/mail data collection, conducted with a national probability sample of U.S. adults from 2020 to 2022. The survey topics focus on reproductive health and family formation. We compare estimates from this survey to those obtained from a face-to-face national survey of population reproductive health: the 2017-2019 National Survey of Family Growth (NSFG). This comparison allows for maximum design complexity, including a complex household screening operation (to identify households with persons aged 18-49). We evaluate the ability of this national web/mail data collection approach to (1) recruit a representative sample of U.S. persons aged 18-49; (2) replicate key survey estimates based on the NSFG, considering expected effects of the COVID-19 pandemic lockdowns and the alternative modes on the estimates; (3) reduce complex sample design effects relative to the NSFG; and (4) reduce the costs per completed survey.

Effect of sampling design on characterizing surface soil fingerprinting properties

Effect of sampling design on characterizing surface soil fingerprinting properties

Effect of training sample size, sampling design and prediction model on soil mapping with proximal sensing data for precision liming

Site-specific estimation of lime requirement requires high-resolution maps of soil organic carbon (SOC), clay and pH. These maps can be generated with digital soil mapping models fitted on covariates observed by proximal soil sensors. However, the quality of the derived maps depends on the applied methodology. We assessed the effects of (i) training sample size (5–100); (ii) sampling design (simple random sampling (SRS), conditioned Latin hypercube sampling (cLHS) and k-means sampling (KM)); and (iii) prediction model (multiple linear regression (MLR) and random forest (RF)) on the prediction performance for the above mentioned three soil properties. The case study is based on conditional geostatistical simulations using 250 soil samples from a 51 ha field in Eastern Germany. Lin’s concordance correlation coefficient (CCC) and root-mean-square error (RMSE) were used to evaluate model performances. Results show that with increasing training sample sizes, relative improvements of RMSE and CCC decreased exponentially. We found the lowest median RMSE values with 100 training observations i.e., 1.73%, 0.21% and 0.3 for clay, SOC and pH, respectively. However, already with a sample size of 10, models of moderate quality (CCC > 0.65) were obtained for all three soil properties. cLHS and KM performed significantly better than SRS. MLR showed lower median RMSE values than RF for SOC and pH for smaller sample sizes, but RF outperformed MLR if at least 25–30 or 75–100 soil samples were used for SOC or pH, respectively. For clay, the median RMSE was lower with RF, regardless of sample size.

SARS-CoV-2 Infection During the First and Second Pandemic Waves in Spain: the ENE-COVID Study.

Objectives. To describe participant characteristics associated with severe acute respiratory syndrome coronavirus 2 infection in Spain's first 2 COVID-19 waves per the Spanish National Seroepidemiological Survey of SARS-CoV-2 Infection (ENE-COVID). Methods. A representative cohort of the noninstitutionalized Spanish population, selected through stratified 2-stage sampling, answered a questionnaire and received point-of-care testing April to June 2020 (first wave: n = 68 287); previously seronegative participants repeated the questionnaire and test November 2020 (second wave: n = 44 451). We estimated seropositivity by wave and participant characteristics, accounting for sampling weights, nonresponse, and design effects. Results. We found that 6.0% (95% confidence interval [CI] = 5.7%, 6.4%) of Spain's population was infected by June and 3.8% (95% CI = 3.5%, 4.1%) more by November 2020. Both genders were equally affected. Seroprevalence decreased with age in adults 20 years and older in the second wave; socioeconomic differences increased. Health care workers were affected at 11.1% (95% CI = 9.0%, 13.6%) and 6.1% (95% CI = 4.4%, 8.5%) in the first and second waves, respectively. Living with an infected person increased infection risk to 22.1% (95% CI = 18.9%, 25.6%) in the first and 35.0% (95% CI = 30.8%, 39.4%) in the second wave. Conclusions. ENE-COVID characterized the first 2 pandemic waves, when information from surveillance systems was incomplete. (Am J Public Health. 2023;113(5):533-544. https://doi.org/10.2105/AJPH.2023.307233).

Effect of sampling design on estimation of phylogenetic diversity metrics of fish community.

Phylogenetic diversity has been widely used to explore diversity patterns and assess processes governing the species composition in community. The estimates of many metrics depend on high-quality data collected from well-designed sampling surveys. However, knowledge of impacts of sampling design on estimation of phylogenetic diversity metrics remains unclear. This study is aim to evaluate the influence of sampling design on phylogenetic diversity metrics estimation of fish community. Simple random sampling (SRS), systematic sampling (SS) and stratified random sampling (StRS) with different sampling intensities were chosen and mean pairwise distances (MPD), mean nearest taxon distance (MNTD), phylogenetic diversity (PD), phylogenetic species variability (PSV), phylogenetic species evenness (PSE) and phylogenetic species richness (PSR) were selected. SRS and StRS showed similar impact on phylogenetic diversity indices estimation and performed relatively well for collecting data to estimate phylogenetic diversity. The accuracy and precision of the estimation increased with sampling intensity under SRS and StRS except SS. MNTD was the only metric not underestimated in four seasons. Metrics strongly influenced by species richness were underestimated when sampling intensity was insufficient. MPD, PSV and PSE showed an obvious seasonal change, which was due to the seasonal differences in fish species composition. In cases where under-sampling is suspected or logistically unavoidable, phylogenetic diversity metrics that are relatively insensitive to sampling design (e.g., MPD and PSV) should be prioritized, especially for exploring the temporal variation infish community. This study reveals it is indispensable to evaluate sampling design when estimating phylogenetic diversity metrics, especially those indices susceptible to species richness.

Physical Environmental Hazards and Major Depression in Older U.S. Adults: The Health and Retirement Study

Background and Aim. Major depression is a leading cause of disability. Although much research focuses on individual risk factors and interventions, environmental risk factors may also play an important role in major depression. We quantified associations of road noise and light at night with prevalent major depression. Methods. We used data from the 2008-2016 biennial interviews of the Health and Retirement Study, a nationally-representative cohort of older adults in the United States. Major depression status was determined using the Composite International Diagnostic Interview Short Form, with a cutoff of ≥5. We estimated average road noise and light at night for the year prior to each interview at participant residences using the Department of Transportation Road Noise Model and satellite imagery from the Defense Meteorological Satellite Program, respectively. To estimate prevalence ratios, we used a modified Poisson regression model that accounted for sampling weights and design effects, adjusted for demographic characteristics, individual and neighborhood SES, time, co-exposures, and geographic area. Results. There were 23,874 participants with complete data. The mean baseline age was 63 years, 53% were female, 74% were Non-Hispanic White, and 15% ever reported major depression. Compared to the lowest exposure quartiles, those in the highest quartiles of noise and light exposure were more likely to be Black, have less than a high school education and reside in an urban area. In fully adjusted models, interquartile differences in both noise and light were suggestively associated with an approximately 10% greater prevalence of major depression (PR per IQR of noise: 1.08, 95% CI: 0.98, 1.20; and light: 1.12, 95%CI: 0.96, 1.31). Conclusion. Exposure to noise and light at night may increase depression in older adults, possibly by disrupting health sleep. Modifying physical hazards in the environment may be a way to reduce the burden of depression. Keywords. Noise, Light, Depression

The second national anti-tuberculosis drug resistance survey in Tanzania, 2017-2018.

To determine the levels and patterns of resistance to first- and second-line anti-tuberculosis (TB) drugs among new and previously treated sputum smear positive pulmonary TB (PTB) patients. We conducted a nationally representative cross-sectional facility-based survey in June 2017-July 2018 involving 45 clusters selected based on probability proportional to size. The survey aimed to determine the prevalence of anti-TB drug resistance and associated risk factors among smear positive PTB patients in Tanzania. Sputum samples were examined using smear microscopy, Xpert MTB/RIF, culture and drug susceptibility testing (DST). Logistic regression was used to account for missing data and sampling design effects on the estimates and their standard errors. We enrolled 1557 TB patients, including 1408 (90.4%) newly diagnosed and 149 (9.6%) previously treated patients. The prevalence of multidrug-resistant TB (MDR-TB) was 0.85% [95% confidence interval (CI): 0.4-1.3] among new cases and 4.6% (95% CI: 1.1-8.2) among previously treated cases. The prevalence of Mycobacterium tuberculosis strains resistant to any of the four first-line anti-TB drugs (isoniazid, rifampicin, streptomycin and ethambutol) was 1.7% among new TB patients and 6.5% among those previously treated. Drug resistance to all first-line drugs was similar (0.1%) in new and previously treated patients. None of the isolates displayed poly-resistance or extensively drug-resistant TB (XDR-TB). The only risk factor for MDR-TB was history of previous TB treatment (odds ratio=5.7, 95% CI: 1.9-17.2). The burden of MDR-TB in the country was relatively low with no evidence of XDR-TB. Given the overall small number of MDR-TB cases in this survey, it will be beneficial focusing efforts on intensified case detection including universal DST.

The Effect of Sampling Design on Data Precision and Practicality in Mobile-LiDAR-based Forest Inventories: Case Study from Rize Senyuva

Bu çalışmada, mobil lazer tarama (LiDAR) sistemleriyle gerçekleştirilecek orman envanterleri için optimum örnekleme tasarımının ortaya koyulması amaçlanmıştır. Çalışma, Şenyuva Orman İşletme Şefliği’ndeki saf Ladin sahasında, geleneksel envanter yöntemleriyle ölçülen örnek alanlara dayalı olarak yürütülmüştür. Aynı saha, mobil LiDAR ile farklı şekillerde taranmış ve üretilen nokta bulutları farklı büyüklüklerde kesilerek, daha önce ölçülen meşcere parametrelerinin değerleri bu yöntemle de belirlenmiştir. Geleneksel yöntemle (çapölçerle) tam alanda (1.834,4 m2) ölçülen göğüs çapı ile LiDAR verisi üzerinden belirlenen göğüs çapı değerleri arasında önemli bir fark tespit edilmemiştir (p>0.05). İki veri setinin ortalamaları arasındaki fark 1 cm’nin (~%2) altındadır. Birim alandaki ağaç sayıları birebir aynı bulunmuştur. Hâkim ağaç boyu ise 1 m’nin (~%1,5) altında hatayla belirlenebilmiştir. Sonuçta, farklı koşullarda çalışan uygulayıcılar için farklı reçeteler geliştirilmiştir. Genel itibariyle, örneklemede yaklaşık ±%10’luk hata payı kabul görüyorsa; saha dışarıdan tarandıktan sonra verinin 400 m2’lik daire şeklinde kesilerek analiz edilmesi önerilebilir. Daha hassas araştırmalarda ise saha dışarıdan tarandıktan sonra tam alanda analiz gerçekleştirilmelidir. Eğer 1 ha’dan büyük sahalarda çalışılacaksa, meşcerenin içeriden taranması zorunlu hale gelmektedir.

HPV vaccination in women living with HIV in the United States (133)

HPV vaccination in women living with HIV in the United States (133)

Seroepidemiology of Crimean-Congo Haemorrhagic Fever among cattle in Cameroon: Implications from a One Health perspective.

BackgroundCrimean-Congo Haemorrhagic Fever (CCHF) is a tick-borne viral zoonotic disease distributed across several continents and recognized as an ongoing health threat. In humans, the infection can progress to a severe disease with high fatality, raising public health concerns due to the limited prophylactic and therapeutic options available. Animal species, clinically unaffected by the virus, serve as viral reservoirs and amplifier hosts, and can be a valuable tool for surveillance. Little is known about the occurrence and prevalence of Crimean-Congo Haemorrhagic Fever Virus (CCHFV) in Cameroon. Knowledge on CCHFV exposure and the factors associated with its presence in sentinel species are a valuable resource to better understand transmission dynamics and assess local risks for zoonotic disease emergence.Methods and findingsWe conducted a CCHFV serological survey and risk factor analysis for animal level seropositivity in pastoral and dairy cattle in the North West Region (NWR) and the Vina Division (VD) of the Adamawa Region in Cameroon. Seroprevalence estimates were adjusted for sampling design-effects and test performance. In addition, explanatory multivariable logistic regression mixed-effects models were fit to estimate the effect of animal characteristics, husbandry practices, risk contacts and ecological features on the serological status of pastoral cattle. The overall seroprevalence was 56.0% (95% CI 53.5–58.6) and 6.7% (95% CI 2.6–16.1) among pastoral and dairy cattle, respectively. Animals going on transhumance had twice the odds of being seropositive (OR 2.0, 95% CI 1.1–3.8), indicating that animal movements could be implicated in disease expansion. From an ecological perspective, absolute humidity (OR 0.6, 95% CI 0.4–0.9) and shrub density (OR 2.1, 95% CI 1.4–3.2) were associated with seropositivity, which suggests an underlying viral dynamic connecting vertebrate host and ticks in a complex transmission network.ConclusionsThis study demonstrated high seroprevalence levels of CCHFV antibodies in cattle in Cameroon indicating a potential risk to human populations. However, current understanding of the underlying dynamics of CCHFV locally and the real risk for human populations is incomplete. Further studies designed using a One Health approach are required to improve local knowledge of the disease, host interactions and environmental risk factors. This information is crucial to better project the risks for human populations located in CCHFV-suitable ecological niches.

Brumadinho Health Project: methodological aspects and epidemiological profile of participants in the cohort baseline.

To present the methodological aspects of the Brumadinho Health Project and to describe the epidemiological profile of participants in the baseline cohort. Prospective, population-based cohort study in a representative sample of residents (aged 12 and over) of Brumadinho, Minas Gerais, after a mining tailings dam failure. Information for the baseline was collected in 2021, two years after the mining tailings dam collapsed, including sociodemographic, health and service use aspects, among others. Prevalence estimates of health outcomes were described in Brumadinho, as well as in the Metropolitan Region of Belo Horizonte and Minas Gerais, using data from the 2019 National Health Survey. All analyses were performed in the software Stata 17.0, considering the sampling weights and design effect. 3,080 (86.4%) residents participated in the study, most of them being females (56.7%) and with a mean age of 46.1 years. The diseases more frequently reported were arterial hypertension (30.1%), high cholesterol (23.1%) and depression (22.5%), similarly to what was observed in the Metropolitan Region of Belo Horizonte and Minas Gerais, although the prevalence in Brumadinho was higher. At least one medical appointment and one hospitalization occurred in 75.2% and 9.4% of residents in the past year, respectively. It is important to monitor health, physical and mental conditions of residents after the occurrence of a disaster of this magnitude. This information can contribute with risk management of these processes, not only in the affected municipality, but also in other areas where populations are at risk of major disasters.

Cross-classes domain inference with network sampling for natural resource inventory

There are two distinct types of domains, design- and cross-classes domains , with the former extensively studied under the topic of small-area estimation. In natural resource inventory, however, most classes listed in the condition tables of national inventory programs are characterized as cross-classes domains, such as vegetation type, productivity class, and age class. To date, challenges remain active for inventorying cross-classes domains because these domains are usually of unknown sampling frame and spatial distribution with the result that inference relies on population-level as opposed to domain-level sampling. Multiple challenges are noteworthy: (1) efficient sampling strategies are difficult to develop because of little priori information about the target domain; (2) domain inference relies on a sample designed for the population, so within-domain sample sizes could be too small to support a precise estimation; and (3) increasing sample size for the population does not ensure an increase to the domain, so actual sample size for a target domain remains highly uncertain, particularly for small domains. In this paper, we introduce a design-based generalized systematic adaptive cluster sampling (GSACS) for inventorying cross-classes domains. Design-unbiased Hansen-Hurwitz and Horvitz-Thompson estimators are derived for domain totals and compared within GSACS and with systematic sampling (SYS). Comprehensive Monte Carlo simulations show that (1) GSACS Hansen-Hurwitz and Horvitz-Thompson estimators are unbiased and equally efficient, whereas the latter outperforms the former for supporting a sample of size one; (2) SYS is a special case of GSACS while the latter outperforms the former in terms of increased efficiency and reduced intensity; (3) GSACS Horvitz-Thompson variance estimator is design-unbiased for a single SYS sample; and (4) rules-of-thumb summarized with respect to sampling design and spatial effect improve precision. Because inventorying a mini domain is analogous to inventorying a rare variable, alternative network sampling procedures are also readily available for inventorying cross-classes domains. ⁃ Generalized systematic adaptive cluster sampling devised for inventorying cross-classes domains ⁃ Unbiased Hansen-Hurwitz and Horvitz-Thompson estimators derived for domain estimation ⁃ Analytical and empirical comparisons made within GSACS and with systematic sampling ⁃ GSACS provides estimates identical to SS using observations fewer on the order of 20%–40% ⁃ Network sampling procedures including GSACS are efficient for inventorying small domains

Sampling plan and methodological aspects: a household healthcare survey in Piauí.

ABSTRACTOBJECTIVE To describe the methodological aspects of the Piauí home healthcare survey (ISAD-PI) and assess the relation between sampling plan, precision, and design effects, assuming that population health surveys are relevant instruments for health monitoring.METHODS ISAD-PI was a population-based cross-sectional study that analyzed the living conditions and health status of the population residing in urban areas in the municipalities of Teresina and Picos, in Piauí. Sampling was carried out by conglomerates in two stages: Primary Sampling Units and households. To calculate the sample size, we considered the stratification of the population in both cities, according to the age of the individuals. We evaluated the “non-response” rate (NRR) and estimated the proportions according to sex and age, as well as the prevalence of social determinants of health in order to assess the compliance of the sampling plan. Analyses related to the precision according to the coefficient of proportion variation of the standard error (Cv-pˆ) and the design effect (deff). Cv-pˆ less than 20% and deff less than 1.5 were considered adequate. The total NRR of households was 38.2% in Teresina and 38.3% in Picos. We carried out twenty-four proportion estimates in relation to age and sex and 48 prevalence of social and health determinants estimates, totaling 72 estimates. Among them, 71 had Cv-pˆ less than 20% and 61 had deff less than or equal to 1.5.CONCLUSION Data generated from the ISAD-PI may contribute to the assessment of health and morbidity conditions in the population. Furthermore, methodological aspects employed in this research may serve as a basis for studies carried out in other cities in Brazil.

Identifying individuals with multiple non-communicable disease risk factors in Kenya: a latent class analysis

ObjectivesNon-communicable diseases (NCDs) are the leading causes of death globally. In Kenya, the number of deaths resulting from NCDs is projected to surpass malaria and tuberculosis by 2030. Studies in Kenya show increasing NCDs; the aim of the present study is to examine the clustering of NCDs and risk factors in Kenya. Study designThis is a cross-sectional study using data from the 2015 Kenya STEPwise Survey. MethodsThis study examined relationships between NCDs (e.g. obesity, hypertension and diabetes) and health behaviours (e.g. sedentary activity, and fruit and vegetable consumption). Survey probability weights, which estimated the sampling design effect, were applied to consider the sampling units, and stratifications were used during sampling so that the results could be generalisable to the national adult Kenyan population. In total, 4350 adults were included in the study sample. ResultsOverall, 24.43% of participants were classified as having hypertension, 1.88% as having type 2 diabetes, and 27.94% were classified as being overweight or obese. The best-fit model was a four-class solution. Class 1 is best described as ‘young with high NCD risk’ and had the highest sedentary activity. Class 2 is best described as ‘poor rural with lower NCD risk’ with a high chance of smoking and alcohol consumption. Class 3 is best described as ‘rural with high NCD risk’ and had the highest fruit and vegetable consumption. Class 4 is best described as ‘wealthy young urban dwellers with high NCD risk’ with a high chance of alcohol consumption and smoking. Individuals in Class 4 had the highest chance (40%) of being overweight/obese, a 2% chance of type 2 diabetes and a 23% chance of having hypertension. ConclusionsNCDs are clustered in groups with high-risk behaviours. The group with the highest chance of having NCDs also had the highest chance of engaging in high-risk behaviours. The findings of this study suggest that smoking and alcohol consumption increase NCD risk in rural areas. Tailored and targeted interventions are needed to curb the increasing NCD prevalence in Kenya.

Onshore biological sampling of landings by species and size category within auction sites can be more efficient than trip-based concurrent sampling

Abstract Since 1996, the European Union has required that fishery products of 35 fish species or groups of species, including crustaceans and cephalopods, be graded before being landed on the basis of commercial size-categories. A multi-stage sampling scheme to estimate catch-at-length and catch-at-age compositions of total annual landings has been conducted in Portuguese waters since 2009. All species and their size-categories are sampled concurrently from random trips within representatively selected site-days annually. In an effort to improve cost-efficiency of the biological catch sampling of commercial landings in Portugal, a size- category sampling scheme was tested during 2017, taking advantage of the stratification of horse mackerel landings, induced by the mandatory grading of landings by size-category. The total number of site-days (primary sampling units), fish boxes sampled, and specimens of horse mackerel sampled for length and age across strata during the pilot study were 30%, 57%, and 12% of the sampling carried out under the standard concurrent scheme, respectively. The assignment of horse mackerel length to the six commercial size-categories was highly consistent across site-days. Our study shows that the concurrent sampling, where trips are subsampled within site-days, could miss some size-categories by chance, resulting in poor estimates of catch-at-length especially for small fish present at the market during a site-day. The size-category sampling scheme ensured subsamples of fish from all size-categories within a site-day, and achieved data on catch-at-age that are fit-for purpose at 24% of the total cost for the standard concurrent sampling. The effect of sampling designs and sample sizes on horse mackerel stock assessment outputs showed that the onshore biological sampling to estimate catch-at-age compositions for stock assessment is optimized with the size-category sampling scheme. Pilot field experiments should be carried out for selected species that evidence consistent size grading among commercial categories.

Want to model a species niche? A step-by-step guideline on correlative ecological niche modelling

The use of correlative ecological niche models has highly increased in the last decade. Despite all literature and textbooks in this field, few practical guidelines exist on the correct application of these techniques. We present here a step-by-step guideline explaining best practices for calculating correlative ecological niche models considering their conceptual and statistical assumptions and limitations. We divided the modelling process into four stages: 1) data collection and preparation; 2) model calculation; 3) model evaluation and validation; 4) and model application. Based on ecological niche theory, we review the concepts of ecological niche and how they can be modelled; classes of correlative models; modelling software; selection of study area; data sources for species records and environmental variables; types of species records and their influence on correlative models; errors in species records; minimum number of species records and environmental variables; effects of prevalence, sampling design, biases, and collinearity between variables; model calculation; model projection to different scenarios in time and space; ensemble modelling; model validation; classification, discrimination and calibration metrics; calculation of null models; analysis of model results; and model thresholding. This guideline is expected to help potential users to obtain better results when using correlative ecological niche models.

Spatial extrapolation of topographic models for mapping soil organic carbon using local samples

Accurate prediction of soil organic carbon (SOC) at fine scales is frequently limited by data availability but fine-scale maps of soil properties are often of great importance for local agricultural land management and watershed processes assessment. Terrain attributes derived from elevation have been commonly used predictor variables in SOC mapping. The aim of this study is to map SOC density (0–30 cm) using hybrid methods that combine regional topographic models and a residual kriging procedure using local samples from an area of interest and evaluate the effects of different sampling designs for residual kriging on final map quality. The regional topographic model was developed based on light detection and ranging (lidar)-derived topographic covariates using random forest to map SOC density on a cropland site intensively sampled for SOC (n = 224). To test ability to generate a local calibration of this regional model, a second intensively sampled cropland site (n = 225) was used to generate a reference SOC density map where residual kriging was implemented involving different numbers of local samples (n = 10–255) and different sample selection schemes: stratified random sampling (SRS) and spatial coverage sampling with close-pairs inclusions (SCS+). To generate a population of local sample sets, 500 repeats were conducted for each scheme from the reference map. By analyzing the distribution of relative mean error (RME) of the overall map for the second site with repeated selection of local samples, we found a smaller variation in RME for SCS+ than that for SRS for a given sample size. However, the E(RME) estimates for SCS+ slightly varied with number of local samples used while the E(RME) for SRS was stabilized at zero over that range in local sampling intensity. There was negligible difference between the distributions of RME for the two sampling schemes at individual points, except a finding of somewhat lower RME for SCS+ based point estimates. These results indicate that SCS+ can generate smaller variation in overall map quality while SRS can provide unbiased mean estimate for an area of interest. This study demonstrates the utility of extrapolating a regional topographic model by use of a limited number of new samples for local calibration and provides insight for practitioners who are interested in mapping soil carbon distribution in sparsely sampled areas of interest.

Seroprevalence of anti-SARS-CoV-2 antibodies in Iquitos, Peru in July and August, 2020: a population-based study

BackgroundDetection of anti-SARS-CoV-2 antibodies among people at risk of infection is crucial for understanding both the past transmission of COVID-19 and vulnerability of the population to continuing transmission and, when done serially, the intensity of ongoing transmission over an interval in a community. We aimed to estimate the seroprevalence of COVID-19 in a representative population-based cohort in Iquitos, one of the regions with the highest mortality rates from COVID-19 in Peru, where a devastating number of cases occurred in March, 2020.MethodsWe did a population-based study of SARS-CoV-2 transmission in Iquitos at two timepoints: July 13–18, 2020 (baseline), and Aug 13–18, 2020 (1-month follow-up). We obtained a geographically stratified representative sample of the city population using the 2017 census data, which was updated on Jan 20, 2020. We included people who were inhabitants of Iquitos since COVID-19 was identified in Peru (March 6, 2020) or earlier. We excluded people living in institutions, people receiving any pharmacological treatment for COVID-19, people with any contraindication for phlebotomy, and health workers or individuals living with an active health worker. We tested each participant for IgG and IgM anti-SARS-CoV-2 antibodies using the COVID-19 IgG/IgM Rapid Test (Zhejiang Orient Gene Biotech, China). We used survey analysis methods to estimate seroprevalence accounting for the sampling design effect and test performance characteristics.FindingsWe identified 726 eligible individuals and enrolled a total of 716 participants (99%), distributed across 40 strata (four districts, two sexes, and five age groups). We excluded ten individuals who: did not have consent from a parent or legal representative (n=3), had moved to Iquitos after March 6, 2020 (n=3), were in transit (n=2), or had respiratory symptoms (n=1). After adjusting for the study sampling effects and sensitivity and specificity of the test, we estimated a seroprevalence of 70% (95% CI 67–73) at baseline and 66% (95% CI 62–70) at 1 month of follow-up, with a test-retest positivity of 65% (95% CI 61–68), and an incidence of new exposures of 2% (95% CI 1–3). We observed significant differences in the seroprevalence between age groups, with participants aged 18–29 years having lower seroprevalence than those aged younger than 12 years (prevalence ratio 0·85 [95% CI 0·73–0·98]; p=0·029).InterpretationAfter the first epidemic peak, Iquitos had one of the highest rates of seroprevalence of anti-SARS-CoV-2 antibodies worldwide. Nevertheless, the city experienced a second wave starting in January, 2021, probably due to the emergence of the SARS-CoV-2 P1 variant, which has shown higher transmissibility and reinfection rates.FundingDirección Regional de Salud de Loreto (DIRESA), Loreto, Peru.TranslationFor the Spanish translation of the abstract see Supplementary Materials section.

Effects of non-representative sampling design on multi-scale habitat models: flammulated owls in the Rocky Mountains.

Sampling bias and autocorrelation can lead to erroneous estimates of habitat selection, model overfitting and elevated omission rates. We developed a multi-scale habitat suitability model of the flammulated owl (Psiloscops flammeolus) in the Northern Rocky Mountains based on extensive but spatially clustered survey data, and then used simulations to evaluate the effects of spatially non-representative and spatially representative sampling strategies on model performance and predictions. Our hypothesis was that models trained with spatially non-representative simulated datasets would suffer from bias in parameter estimates, and would show lower predictive performance. The models trained with the spatially representative simulated datasets greatly outperformed the models trained with the spatially non-representative simulated datasets judged on standard metrics of model performance. However, the spatially non-representative models produced superior predictions based on their ability to identify the correct spatial scales, covariates, signs and magnitudes of the species-environment relationships, when compared to the spatially representative models. Thus, it is likely that representative spatial sampling across a broad range of environmental gradients also resulted in over-dispersion of sampling data, with a higher proportion of samples falling in areas of low probability of presence, leading to lower ability to resolve the relationships between species presence-absence and environmental covariates. In contrast, the spatially non-representative sampling, by concentrating sampling along environmental gradients that are characterized by higher probability of presence of the modelled species, produced predictions that, while seeming to be weaker based on standard measures of model performance (e.g., AUC, Kappa, PCC), greatly outperformed the spatially representative models based on measures of true model prediction (e.g., correctly describing the actual spatial scales, direction and strength of species-environment relationships). Further work using simulation approaches is warranted to more fully evaluate the ability of species distribution modelling techniques to correctly identify scales, driving covariates, signs and magnitudes of relationships between species presence-absence patterns, and environmental covariates.

Assessing the Effect of Training Sampling Design on the Performance of Machine Learning Classifiers for Land Cover Mapping Using Multi-Temporal Remote Sensing Data and Google Earth Engine

Machine learning classifiers are being increasingly used nowadays for Land Use and Land Cover (LULC) mapping from remote sensing images. However, arriving at the right choice of classifier requires understanding the main factors influencing their performance. The present study investigated firstly the effect of training sampling design on the classification results obtained by Random Forest (RF) classifier and, secondly, it compared its performance with other machine learning classifiers for LULC mapping using multi-temporal satellite remote sensing data and the Google Earth Engine (GEE) platform. We evaluated the impact of three sampling methods, namely Stratified Equal Random Sampling (SRS(Eq)), Stratified Proportional Random Sampling (SRS(Prop)), and Stratified Systematic Sampling (SSS) upon the classification results obtained by the RF trained LULC model. Our results showed that the SRS(Prop) method favors major classes while achieving good overall accuracy. The SRS(Eq) method provides good class-level accuracies, even for minority classes, whereas the SSS method performs well for areas with large intra-class variability. Toward evaluating the performance of machine learning classifiers, RF outperformed Classification and Regression Trees (CART), Support Vector Machine (SVM), and Relevance Vector Machine (RVM) with a >95% confidence level. The performance of CART and SVM classifiers were found to be similar. RVM achieved good classification results with a limited number of training samples.