China
Africa
Explore chapters and articles related to this topic
Assessing the relationship between exercise and employee mental health: methodological concerns
Published in John Kerr, Amanda Griffiths, Tom Cox, Workplace Health, Employee Fitness and Exercise, 2020
Steve M. Jex, Deanne A. Heinisch
Statistical power is the extent to which a study is capable of detecting true effects when they exist. According to Cohen (1969), statistical power is primarily a function of the size of the effect that a researcher is trying to detect, measurement reliability and sample size. Assuming that a researcher is using reliable measures, relatively small sample sizes are sufficient to detect large effects. Conversely, large sample sizes are needed to detect small effects. According to Cohen (1992), researchers in many areas of psychology have neglected the issue of statistical power. Furthermore, such neglect on the part of researchers comes with the increased danger of drawing erroneous conclusions from research results. For example, a researcher may conduct an experiment and conclude that the experimental treatment has no effect. The real problem, however, may be that the effect is small and statistical power was lacking.
Involving older adults in design research
Published in Sara J. Czaja, Walter R. Boot, Neil Charness, Wendy A. Rogers, Designing for Older Adults, 2019
Sara J. Czaja, Walter R. Boot, Neil Charness, Wendy A. Rogers
In larger efficacy or effectiveness trials, statistical power is a critical issue as it impacts the confidence that can be placed in the findings of the study. Statistical power is the extent to which the study can detect the difference between two groups and is a function of three factors: the criterion established for statistical significance (alpha level, typically set at .05), the difference that exists between the groups (effect size), and the sample size. Various algorithms and software programs are available to help calculate statistical power and derive needed sample size. Calculation of the appropriate sample size must occur prior to the beginning of the study for planning purposes. In addition to statistical power, the number of participants that will be required impacts the recruitment strategy, staffing requirements, budget, and timeline.
Hypothesis testing
Published in Dev P. Chakraborty, Observer Performance Methods for Diagnostic Imaging, 2017
In most statistical books, the subject of hypothesis testing is demonstrated in different (i.e., non-ROC) contexts. That is to be expected since this field is a small subspecialty of statistics (Prof. Howard E. Rockette, private communication, ca. 2002). Since this book is about ROC analysis, the author decided to use a demonstration using ROC analysis. Using a data simulator, one is allowed to cheat by conducting a very large number of simulations to estimate the population AUC and standard deviation under the null hypothesis. This permitted us to explore the related concepts of Type-I and Type-II errors within the context of ROC analysis. Ideally, both errors should be zero, but the nature of statistics leads one to two compromises. Usually one accepts a Type-I error capped at 5% and a Type-II error capped at 20%. These translate to α= 0.05 and statistical power = 80%. The dependence of statistical power on α, the numbers of cases, and the effect size was explored. Statistical power increases with the effect size, it increases with α, and it increases with the sample size (numbers of cases).
Environmental behaviour and choice of sustainable travel mode in urban areas: comparative evidence from commuters in Asian cities
Published in Production Planning & Control, 2020
Junya Kumagai, Shunsuke Managi
The sample size of our analytical data is relatively smaller than common empirical studies. Generally, a smaller sample size causes lower statistical power, which means that the results are more likely to show no significant difference even if there is difference in reality (Björklund and Swärdh 2017; Vergouwe et al. 2005). In addition, small sample size in logistic regression may overestimate the effects of explanatory variables. Long (1997) suggested that the sample size smaller than 100 is not enough to maximum likelihood estimation, while sample size larger than 500 seem adequate. Moreover, additional observations per each additional unknown parameter are needed. Nemes et al. (2009) argued that the necessary sample size varies depending on the data structure, so that Long’s numerical suggestion may not be all true every time.
Force production during maximal front crawl tethered swimming: exploring bilateral asymmetries and differences between breathing and non-breathing conditions
Published in Sports Biomechanics, 2021
Stelios G. Psycharakis, Helen Soultanakis, José María González Ravé, Giorgios P. Paradisis
The present study has some limitations that need to be taken into consideration when interpreting the results. First, due to between-gender differences in most variables, subsequent analysis had to be performed separately for each gender. This resulted in smaller sample sizes, which reduce statistical power. Thus, some of the variables in the present study may also show significant changes if the study is repeated with larger sample sizes. Second, our group contained university level swimmers, whose 50 m SB was circa 79% of the world record, and it is therefore not known if the same patterns would exist in elite national and international level swimmers. Third, we chose to compare the force between dominant and non-dominant arms while the swimmers were also performing maximal kicking. This comparison assumes that the maximal kicking between these two conditions would be nearly identical. This is a relatively reasonable assumption for short maximal bouts without breathing; for example, the Fmin in these one-arm trials would most probably be recorded during the arm recovery (when the only propulsion comes from the kicking actions), and the fact that the Fmin values showed no significant differences between conditions suggests that kicking actions were broadly similar. Nevertheless, it is not possible to confirm if there were no differences for any of the swimmers in the propulsive effect of the kicking actions in the two conditions. Performing the trials without any kicking could have been an option, but pilot testing indicated that the lack of propulsive continuity and the demands of tethered swimming would have made this task very difficult without sacrificing body position and normal swimming technique.
Semen quality and sperm DNA damage associa –revised – final-finalted with oxidative stress in relation to exposure to polycyclic aromatic hydrocarbons
Published in Journal of Environmental Science and Health, Part A, 2018
Hueiwang Anna C. Jeng, Wen Y. Lin, Mu R. Chao, Wen Y. Lin, Chih H. Pan
This study included limitations. One limitation was the small sample sizes. Although a power analysis was conducted to address the minimal sample size, it is recommended to increase the sample size to achieve a higher statistical power. Also, a larger sample size could allow stratification of human subjects into subgroups and further assess whether smoking and alcohol consumption affect sperm DNA damage. The second weakness was the lack of knowledge about any prior chemical exposure of the workers before employment at the current company. The company didn’t have records on the history of prior chemical exposure. Our questionnaire did include work history that briefly revealed the length of employment at the current job and prior chemical exposure, particularly metals and volatile organic compounds. We did exclude human subjects who reported more than a 5-year occupational exposure to volatile organic compounds and metals at other jobs. The third weakness was related to a mixture of coke oven emissions, including coal tar, volatiles and PAHs and metals. We did monitor metals in urine from the workers.[41] Also, we excluded the workers with prior exposure to volatile organic compounds employed at other companies. We used urinary 1-OHP concentrations as the biomarkers to depict internal exposure doses of PAHs.[8,9,12–15] We did monitor 16 species of PAHs to ensure that the exposed group had significantly higher PAHs exposure than the control. 21,22] A follow-up study with a larger sample size could assess whether specific PAH species are correlated to sperm DNA damage.