China
Africa
Explore chapters and articles related to this topic
Empirical Likelihood
Published in Albert Vexler, Alan D. Hutson, Xiwei Chen, Statistical Testing Strategies in the Health Sciences, 2017
Albert Vexler, Alan D. Hutson, Xiwei Chen
Lee and Young (1999) showed that higher-order correction to the confidence interval can be obtained by a variety of techniques, including asymptotic adjustment to the chi-square percentiles, bootstrapping the nonparametric likelihood ratio, and direct substitution of truncated asymptotic expansions for the confidence interval endpoints. In a smooth-function model setting, the authors proposed the construction of two-sided nonparametric confidence intervals based on Stein’s least favorable family. The approach shares the same asymptotic properties and the advantages of empirical likelihood, such as having a data-driven shape. In addition, the method does not require nested levels of bootstrap sampling, and thus is computationally less intensive. Based on the method, one may propose and analyze asymptotic and bootstrapping techniques as a way of reducing coverage error to levels. Based on a simulation study, the coverage error was shown to be substantially reduced by simple analytic adjustment of the non-parametric likelihood interval. Moreover, it was suggested that bootstrapping the distribution of the nonparametric likelihood ratio results in very desirable coverage accuracy.
Surveys, longitudinal, cross-sectional and trend studies
Published in Louis Cohen, Lawrence Manion, Keith Morrison, Research Methods in Education, 2017
Louis Cohen, Lawrence Manion, Keith Morrison
Dillman et al. (2014) identify four key errors to be avoided in surveys which seek to represent a wider population: coverage error (poor and incomplete representation of the population in the sample). For example, a coverage error might be made if telephone or Internet surveys are used, as not everyone has a telephone (particularly a landline) or access to, and familiarity with, the Internet;sampling error (including inaccurate estimates of the population);non-response error (the difference between a representative result and that obtained from non-response of different individual or groups, i.e. a skewed response); andmeasurement error: inaccurate and unreliable response because of (a) the metrics, scales and units of measurement used; (b) socially desirable responses and respondent acquiescence (the tendency to agree with an interviewer rather than disagree) in face-to-face survey interviews; (c) questionnaire features, for example, length, difficulty, questions asked, complexity, order effects, interviewer effects, survey mode (post, telephone, email, interview, Internet etc.).
Trimmed dual system estimation
Published in Dankmar Böhning, Peter G.M. van der Heijden, John Bunge, Capture-Recapture Methods for the Social and Medical Sciences, 2017
Apart from the countries that use a completely register-based approach, several countries have replaced the census with a statistical population register, and administered coverage surveys to provide the necessary coverage error adjustments. For instance, in Israel (Nirel and Clickman [217]), the population register yields a reliable overall count for the whole county, but coverage error adjustments are necessary primarily to cope with interregional mis-location. Both the UCS and OCS were deployed in census 2008. The UCS has a sampling fraction of 20%, which is larger than usual and alleviates the need for indirect population estimates. The fieldwork and cost are more limited when it comes to the OCS, because erroneous enumeration is not a major concern in the population register.
Evaluation of knowledge and practices about administration and regulations of high alert medications among hospital pharmacists in Pakistan: findings and implications
Published in Current Medical Research and Opinion, 2022
Muhammad Salman, Zia Ul Mustafa, Naureen Shehzadi, Tauqeer Hussain Mallhi, Noman Asif, Yusra Habib Khan, Tahir Mehmood Khan, Khalid Hussain
The current study is accompanied by a few limitations which should be considered while interpreting the results. Firstly, this was a web-based study administered using a non-probability sampling method, thus, the propensities of various errors including coverage error, referral bias and selection bias cannot be disregarded. Secondly, a self-completed questionnaire was used so introspection and reporting bias might be present. Thirdly, our target population was hospital pharmacists working in various hospitals in the Punjab province of Pakistan so the generalizability of the findings is limited to other provinces of the country. Nonetheless, the current study is strengthened by the first of its kind ascertaining the knowledge of pharmacists related to HAMs. We believe our findings will serve to design and implement educational initiatives by the health authorities to improve patient care.
Recruitment of population-based controls for ALS cases from the National ALS Registry
Published in Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration, 2021
Todd M. Bear, Angela M. Malek, Abigail Foulds, Judith Rager, Sarah E. Deperrior, John E. Vena, Theodore C. Larson, Paul Mehta, D. Kevin Horton, Evelyn O. Talbott
Inherent issues recruiting hospital or acquaintance controls related to confounding have limited past studies. Strengths of the current study include identification of population-based controls using a landline sample. As older adults are more likely to have a landline phone, a landline sample is advantageous for studying ALS, which has a mean/median onset age of 51-66 years (7). Additionally, landline samples provide high accuracy for geo-locating as well as appending addresses including zip codes and other demographic characteristics at high rates, and they are more cost effective. However, as the landline sampling frame continues to deteriorate and coverage error grows, new technologies on the cell phone sampling frame are emerging that will allow for similar demographic and geographic targeting. The coverage error associated with choosing to sample from the landline frame is estimated at approximately 30% for those age 65 and over living in wireless only households (8).
Comprehensive assessment of patients with irritable bowel syndrome with constipation and chronic idiopathic constipation using deterministically linked administrative claims and patient-reported data: the Chronic Constipation and IBS-C Treatment and Outcomes Real-World Research Platform (CONTOR)
Published in Journal of Medical Economics, 2020
Douglas C. A. Taylor, Jessica L. Abel, Carolyn Martin, Jalpa A. Doshi, Breanna Essoi, Stephanie Korrer, David S. Reasner, Robyn T. Carson, Alyssa Goolsby Hunter
This study utilized both survey and claims data to address limitations associated with reliance on either data source alone. Limitations of survey data may include sampling error (including the existence of bias due to unmeasured differences in characteristics of survey respondents and nonrespondents), coverage error, and measurement error. Limitations of claims data include that the presence of a claim for a filled prescription does not indicate the medication was consumed or taken as prescribed; OTC medications and those provided as samples by a physician cannot be observed; and the presence of a diagnosis code on a medical claim is not proof of disease, as diagnosis codes may also be coded incorrectly or included as rule-out criteria. These limitations were addressed in part by asking participants to report current treatments, including frequency of use, and by collecting patient-reported OTC and prescription medication use as opposed to OTCs purchased or prescriptions filled. Use of diagnosis codes was also complicated by lack of IBS subtypes using ICD-9-CM codes for sample identification. This limitation was mitigated by a variety of approaches, including use of patient-reported health care provider diagnoses and modified Rome III criteria.