Excerpts on Sampling: Bias in Sampling

Abstract Background The two main primate groups recorded throughout the European Miocene, hominoids and pliopithecoids, seldom co-occur. Due to both their rarity and insufficiently understood palaeoecology, it is currently unclear whether the infrequent co-occurrence of these groups is due to sampling bias or reflects different ecological preferences. Here we rely on the densely sampled primate-bearing sequence of Abocador de Can Mata (ACM) in Spain to test whether turnovers in primate assemblages are correlated with palaeoenvironmental changes. We reconstruct dietary evolution through time (ca. 12.6–11.4 Ma), and hence climate and habitat, using tooth-wear patterns and carbon and oxygen isotope compositions of enamel of the ubiquitous musk-deer Micromeryx. Results Our results reveal that primate species composition is strongly correlated with distinct environmental phases. Large-bodied hominoids (dryopithecines) are recorded in humid, densely-forested environments on the lowermost portion of the ACM sequence. In contrast, pliopithecoids inhabited less humid, patchy ecosystems, being replaced by dryopithecines and the small-bodied Pliobates toward the top of the series in gallery forests embedded in mosaic environments. Conclusions These results support the view that pliopithecoid primates preferred less humid habitats than hominoids, and reveal that differences in behavioural ecology were the main factor underpinning their rare co-occurrence during the European Miocene. Our findings further support that ACM hominoids, like Miocene apes as a whole, inhabited more seasonal environments than extant apes. Finally, this study highlights the importance of high-resolution, local investigations to complement larger-scale analyses and illustrates that continuous and densely sampled fossiliferous sequences are essential for deciphering the complex interplay between biotic and abiotic factors that shaped past diversity.

Download Full-text

Understanding Users’ Satisfaction Towards Public Transit System in India: A Case-Study of Mumbai

ISPRS International Journal of Geo-Information ◽

10.3390/ijgi10030155 ◽

2021 ◽

Vol 10 (3) ◽

pp. 155

Author(s):

Rahul Das

Keyword(s):

Service Quality ◽

Public Transport ◽

Travel Demand ◽

Resource Constraints ◽

Public Transit ◽

Sampling Bias ◽

Relevant Information ◽

Transit System ◽

Transport Services ◽

Developing Cities

In this work, we present a novel approach to understand the quality of public transit system in resource constrained regions using user-generated contents. With growing urban population, it is getting difficult to manage travel demand in an effective way. This problem is more prevalent in developing cities due to lack of budget and proper surveillance system. Due to resource constraints, developing cities have limited infrastructure to monitor transport services. To improve the quality and patronage of public transit system, authorities often use manual travel surveys. But manual surveys often suffer from quality issues. For example, respondents may not provide all the detailed travel information in a manual travel survey. The survey may have sampling bias. Due to close-ended design (specific questions in the questionnaire), lots of relevant information may not be captured in a manual survey process. To address these issues, we investigated if user-generated contents, for example, Twitter data, can be used to understand service quality in Greater Mumbai in India, which can complement existing manual survey process. To do this, we assumed that, if a tweet is relevant to public transport system and contains negative sentiment, then that tweet expresses user’s dissatisfaction towards the public transport service. Since most of the tweets do not have any explicit geolocation, we also presented a model that does not only extract users’ dissatisfaction towards public transit system but also retrieves the spatial context of dissatisfaction and the potential causes that affect the service quality. It is observed that a Random Forest-based model outperforms other machine learning models, while yielding 0.97 precision and 0.88 F1-score.

Download Full-text

The Road Not Taken: How Early Landscape Learning and Adoption of a Risk-Averse Strategy Influenced Paleoindian Travel Route Decision Making in the Upper Ohio Valley

American Antiquity ◽

10.1017/aaq.2020.96 ◽

2020 ◽

Vol 86 (1) ◽

pp. 133-152

Author(s):

Matthew P. Purtill

Keyword(s):

Time Allocation ◽

Selection Process ◽

Sampling Bias ◽

Ohio River ◽

Risk Averse ◽

Ohio Valley ◽

The Road ◽

Archaeological Data ◽

Risk Sensitive ◽

Path Distance

To evaluate a model of the travel-route selection process for upper Ohio Valley Paleoindian foragers (13,500–11,400 cal BP), this study investigates archaeological data through the theoretical framework of landscape learning and risk-sensitive analysis. Following initial trail placement adjacent to a highly visible escarpment landform, Paleoindians adopted a risk-averse strategy to minimize travel outcome variability when wayfaring between Sandy Springs, a significant Ohio River Paleoindian site, and Upper Mercer–Vanport chert quarries of east-central Ohio. Although a least-cost analysis indicates an optimal route through the lower Scioto Valley, archaeological evidence for this path is lacking. Geomorphic and archaeological data further suggest that site absence in the lower Scioto Valley is not entirely due to sampling bias. Instead, evidence indicates that Paleoindians preferred travel within the Ohio Brush Creek–Baker's Fork valley despite its longer path distance through more rugged, constricted terrain. Potential travel through the lower Scioto Valley hypothesizes high outcome variability due to the stochastic nature of the late Pleistocene hydroregime. In this case, perceived outcome variability appears more influential in determining travel-route decisions among Paleoindians than direct efforts to reduce energy and time allocation.

Download Full-text

Antimicrobial susceptibility testing results from 13 hospitals in Viet Nam: VINARES 2016–2017

Antimicrobial Resistance and Infection Control ◽

10.1186/s13756-021-00937-4 ◽

2021 ◽

Vol 10 (1) ◽

Author(s):

Tien Viet Dung Vu ◽

◽

Marc Choisy ◽

Thi Thuy Nga Do ◽

Van Minh Hoang Nguyen ◽

...

Keyword(s):

Antimicrobial Susceptibility ◽

Susceptibility Testing ◽

Data Entry ◽

Sampling Bias ◽

Antimicrobial Susceptibility Testing ◽

Viet Nam ◽

Salmonella Spp ◽

Invasive Infections ◽

Shigella Spp ◽

Analyse Data

Abstract Objective To analyse data from 2016–17 from a hospital-based antimicrobial resistance surveillance with national coverage in a network of hospitals Viet Nam. Methods We analysed data from 13 hospitals, 3 less than the dataset from the 2012–13 period. Identification and antimicrobial susceptibility testing data from the clinical microbiology laboratories from samples sent in for routine diagnostics were used. Clinical and Laboratory Standards Institute 2018 guidelines were used for antimicrobial susceptibility testing interpretation. WHONET was used for data entry, management and analysis. Results 42,553 deduplicated isolates were included in this analysis; including 30,222 (71%) Gram-negative and 12,331 (29%) Gram-positive bacteria. 8,793 (21%) were from ICUs and 7,439 (18%) isolates were from invasive infections. Escherichia coli and Staphylococcus aureus were the most frequently detected species with 9,092 (21%) and 4,833 isolates (11%), respectively; followed by Klebsiella pneumoniae (3,858 isolates – 9.1%) and Acinetobacter baumannii (3,870 isolates – 9%). Bacteria were mainly isolated from sputum (8,798 isolates – 21%), blood (7,118 isolates – 17%) and urine (5,202 isolates – 12%). Among Gram-positives 3,302/4,515 isolates (73%) of S. aureus were MRSA; 99/290 (34%) of Enterococcus faecium were resistant to vancomycin; and 58% (663/1,136) of Streptococcus pneumoniae proportion were reduced susceptible to penicillin. Among Gram-negatives 59% (4,085/6,953) and 40% (1,186/2,958) of E. coli and K. pneumoniae produced ESBL and 29% (376/1,298) and 11% (961/8,830) were resistant to carbapenems, respectively. 79% (2855/3622) and 45% (1,514/3,376) of Acinetobacter spp. and Pseudomonas aeruginosa were carbapenem resistant, respectively. 88% (804/911) of Haemophilus influenzae were ampicillin resistant and 18/253 (7%) of Salmonella spp. and 7/46 (15%) of Shigella spp. were resistant to fluoroquinolones. The number of isolates from which data were submitted in the 2016–2017 period was twice as high as in 2012–2013. AMR proportions were higher in 2016–2017 for most pathogen-antimicrobial combinations of interest including imipenem-resistant A. baumannii, P. aeruginosa and Enterobacterales. Conclusions The data show alarmingly high and increasing resistant proportions among important organisms in Viet Nam. AMR proportions varied across hospital types and should be interpreted with caution because existing sampling bias and missing information on whether isolates were community or hospital acquired. Affordable and scalable ways to adopt a sample- or case-based approach across the network should be explored and clinical data should be integrated to help provide more accurate inferences of the surveillance data.

Download Full-text

Centrality-friendship paradoxes: when our friends are more important than us

Journal of Complex Networks ◽

10.1093/comnet/cny029 ◽

2018 ◽

Vol 7 (4) ◽

pp. 515-528 ◽

Cited By ~ 4

Author(s):

Desmond J Higham

Keyword(s):

Network Science ◽

Sufficient Conditions ◽

Sampling Bias ◽

Average Degree ◽

Science Literature ◽

Eigenvector Centrality ◽

General Matrix ◽

Theoretical Support ◽

The Social ◽

Katz Centrality

Abstract The friendship paradox states that, on average, our friends have more friends than we do. In network terms, the average degree over the nodes can never exceed the average degree over the neighbours of nodes. This effect, which is a classic example of sampling bias, has attracted much attention in the social science and network science literature, with variations and extensions of the paradox being defined, tested and interpreted. Here, we show that a version of the paradox holds rigorously for eigenvector centrality: on average, our friends are more important than us. We then consider general matrix-function centrality, including Katz centrality, and give sufficient conditions for the paradox to hold. We also discuss which results can be generalized to the cases of directed and weighted edges. In this way, we add theoretical support for a field that has largely been evolving through empirical testing.

Download Full-text