|
|
Meta-analysis of bone tumorous lesions in spinal CT data using convolutional neural networks
Nantl, Ondřej ; Jakubíček, Roman (referee) ; Chmelík, Jiří (advisor)
This bachelor thesis deals with the use of convolutional neural networks in the meta-analysis of bone tumor lesions in CT image data. The theoretical part describes the anatomy and pathology of bone tissue, machine learning, discusses the functionality of convolutional neural networks and summarizes selected existing methods for computer-aided diagnosis of vertebra bone lesions. In the practical part, various types of models using convolutional neural networks were implemented and the networks were trained on an available augmented dataset. Finally, the results of various types of models were statistically evaluated, compared with available articles and discussed.
|
|
|
How Does Peer Socioeconomic Status Affect Academic Achievement? A Meta-Analysis
Markalousová, Tereza ; Havránek, Tomáš (advisor) ; Kantová, Klára (referee)
Using state-of-the-art meta-analysis methods, we investigate the e ect of peer socioeconomic status (SES) on academic achievement. Our data set covers almost 40 years of research, containing 449 estimates from 40 studies. We examine publication bias for the first time in this research area. Various sta- tistical tests reveal the presence of publication selection and suggest a smaller e ect size in comparison to prior findings. Employing Bayesian and frequentist model averaging techniques, we identify factors that systematically influence the magnitude of the estimated e ect. Publication bias exerts the strongest upward e ect, along with the use of parental education and home resources as measures of SES and combined measures of SES. Conversely, variables such as the number of citations, publication status, science test type, and the use of advanced methods exhibit a negative e ect. Keywords Socieconomic status, Meta-analysis, Social class, Academic achievement Title How Does Peer Socioeconomic Status A ect Academic Achievement? A Meta-Analysis
|
|
|
How much does intelligence predict lifetime income? A Meta-Analysis
Nguyenová, Van Anh ; Havránková, Zuzana (advisor) ; Bortnikova, Kseniya (referee)
Despite growing interest and extensive empirical research in economic returns to ability, a consensus regarding the true impact of intelligence on financial outcomes remains elusive. While psychology literature has made e orts to unify divergent findings, economics is yet to produce a comprehensive meta- analysis addressing this issue. Addressing this gap, our thesis utilizes cutting- edge meta-analytic techniques to analyze a unique dataset of 765 estimates drawn from 38 studies, providing a clearer picture of intelligence's impact on income. We uncover a notable positive publication bias, which, after correction, yields a diminished yet statistically significant e ect. Specifically, our results indicate that a standard deviation increase in cognitive ability results in a less than 10% increase in financial outcomes. Leveraging over 30 variables in our Bayesian and frequentist averaging models, we identify key determinants of this e ect, including the data collection year, outcome specifications, methodologi- cal choices, country-specific factors, and the number of estimates reported per study. Additionally, when adjusting for factors such as gender, residential loca- tion, work experience, and family attributes, we observe substantial variations in e ect size. JEL Classification J24, J31, D31, C11...
|
|
|
Standing Tall Pays Off: A Meta-Analysis of Height Premium
Juračková, Martina ; Havránková, Zuzana (advisor) ; Polák, Petr (referee)
As has been demonstrated by empirical research, height is an important physical feature impacting various aspects of the life of an individual. This thesis deals with the relationship between height and income, also referred to as height premium. With the help of modern meta-analytic methods, we aim to quantitatively summarize the empirical evidence on the impact of height on income. After introducing the topic of height premium, data collection and methodological framework, we test for publication bias. The analysis is conducted on 1084 height premium estimates collected from 67 studies. The results of publication bias testing indicate that height premium literature contains positive publication bias which persists even after we control for additional variables capturing study characteristics or, in other words, the heterogeneity of collected estimates. Based on Bayesian Model Averaging results, we conclude that geographical factors, the longitudinal nature of the dataset, restriction of the dataset with respect to gender, or adding a gender control variable into the regression are the most important factors explaining the variability of height premium effects.
|
|
|
The Effect of Face Masks on Covid Transmission: A Meta-Analysis
Lušková, Martina ; Havránková, Zuzana (advisor) ; Bertoli, Paola (referee)
The effect of face masks on Covid-19 transmission is crucial for the health of populations. Nevertheless, its economic consequences cannot be overlooked. To perform a quantitative meta-analysis, 258 estimates from 44 primary studies were collected together with more than 30 variables mirroring the differences among the studies. Publication bias was examined by implementing various statistical tests resulting in mild evidence for the phenomenon. We contribute to other meta-analyses on the topic by employing the Bayesian and Frequen- tist model averaging to identify the drivers behind the heterogeneity of the estimates. The results suggest that temperature, geographical latitude, and panel data structure have a highly statistically significant and positive effect on the risk of transmission associated with mask-wearing. Moreover, a pos- itive effect was identified for healthcare set-up. In contrast, performing an aerosol-generating procedure shifts the risk in the negative direction. JEL Classification I1, I11, I19, Keywords meta-analysis, Covid-19, face masks, pan- demic, Covid-19 transmission, publication bias, Bayesian model averaging Title The Effect of Face Masks on Covid Transmis- sion: A Meta-Analysis
|
|
|
Military expenditure and economic growth: A meta-analysis
Simpartl, Josef ; Havránek, Tomáš (advisor) ; Baxa, Jaromír (referee)
This master's thesis provides a comprehensive meta-analysis inspecting the impact of military expenditures on economic growth. Its objective is to validate several hypotheses using a wide range of studies and the latest methods. The hypotheses included in this thesis are as follows. First, military expenditures reduce economic growth. Second, military expenditures retard growth in less- developed countries. Third, the effect of military expenditures on economic growth is non-linear. The general findings of this thesis support the first hypothesis. This result originates mainly from the significantly negative impact found in recent studies. The thesis also finds indirect evidence in favor of the second hypothesis. Even though the analysis did not find a significant impact on less-developed countries per se, a negative effect associated with African countries provides a persuasive substitute in this regard. Last, this thesis did not find any solid evidence in favor of the non-linearity hypothesis. This is the first time the effect of military expenditures on economic growth was analyzed using the latest methods of meta-analysis and provides robust input into a heated debate within the subfield of peace economics provoked by the recent events. 1
|
|
|
The Causal Effect of Parents' Schooling on Children's Schooling: A Meta-Analysis
Pokorná, Anastasia ; Havránková, Zuzana (advisor) ; Kukačka, Jiří (referee)
One of the topics concerning education's improvement is the intergenerational transmission of education. The main goal of our study is to analyse the causal efect of parents' education on children's education. We collect 387 estimates of the causal efects from 23 studies. Using our dataset and meta-analytic methods, we test for the presence of publication bias in the literature and try to explain the heterogeneity in the results of the primary studies. After correction for publication bias, the mean becomes smaller than the mean reported in the literature and varies from 0.044 to 0.185. For exploring the heterogeneity of the estimates we use Bayesian Model Averaging and Frequentist Model Averaging. Our results suggest that there is a country heterogeneity in the estimates of the causal efect. Moreover, controlling for the size of a household in the original regression is important for the explanation of the diferences in the results of the primary studies. In addition to it, we collect 605 estimates of the non- causal associations from 39 studies and analyse the publication bias and the heterogeneity of results also using this sample. 1
|
|
|
Bank Survival Around the World: A Meta-Analytic Review
Kočenda, Evžen ; Iwasaki, I.
Bank survival is essential to economic growth and development because banks mediate the financing of the economy. A bank’s overall condition is often assessed by a supervisory rating system called CAMELS, an acronym for the components Capital adequacy, Asset quality, Management quality, Earnings, Liquidity, and Sensitivity to market risk. Estimates of the impact of CAMELS components on bank survival vary widely. We perform a meta-synthesis and meta-regression analysis (MRA) using 2120 estimates collected from 50 studies. In the MRA, we account for uncertainty in moderator selection by employing Bayesian model averaging. The results of the synthesis indicate an economically negligible impact of CAMELS variables on bank survival; in addition, the effect of bank-specific, (macro)economic, and market factors is virtually absent. The results of the heterogeneity analysis and publication bias analysis are consistent in terms that they do not find an economically significant impact of the CAMELS variables. Moreover, best practice estimates show a small economic impact of CAMELS components and no impact of other factors. The study concludes that caution should be exercised when using CAMELS rating to predict bank survival or failure.
|
|
|
Essays in Behavioural and Experimental Economics
Matoušek, Jindřich ; Havránek, Tomáš (advisor) ; Miklánek, Tomáš (referee) ; Rachinger, Heiko (referee) ; Gechert, Sebastian (referee)
CHARLES UNIVERSITY FACULTY OF SOCIAL SCIENCES Institute of Economic Studies Essays in Behavioural and Experimental Economics Abstract for Dissertation thesis Author: Mgr. Jindřich Matoušek Study program: Economics and Finance Supervisor: prof. PhDr. Tomáš Havránek, Ph.D. Year of defense: 2022 Abstract The dissertation consists of three papers presenting applications of experimen- tal as well as statistical methods to the topics of behavioural economics. The first paper introduces a series of laboratory experiments in which I apply the experimental methods to a complex decision making problem. The second and third papers present quantitative syntheses of the literature on the classi- cal topics of behavioural economics. The general introduction connects these chapters together. Detailed abstracts for individual papers are presented at the beginning of each chapter. In the first paper, I experimentally examine two complex multi-unit auc- tion mechanisms with an opportunity to communicate and thus collude while comparing these mechanisms in terms of efficiency. Strikingly, allowing for communication increases efficiency in examined auction formats. A cheap-talk collusive agreement resulted in a better allocation compared to the treatments without communication. I hypothesize that complex auction formats makes...
|
|
|
Are Women More Risk-Averse than Men? A Meta-Analysis
Černý, Patrik ; Havránek, Tomáš (advisor) ; Bauer, Michal (referee)
Are women more risk averse than men? While a large portion of economic literature confirms this phenomenon, the link between gender and risk aver- sion has not been found to be consistent. We investigate the difference in risk aversion between genders employing meta-analytic methods on 147 estimates collected from 25 primary studies, converted to partial correlation coefficients. We find positive publication bias in our dataset, suggesting that the reported estimates in primary studies are exaggerated. After adjusting for the publica- tion bias, men do not seem to be more risk-taking than women. In addition, we employ Bayesian Model Averaging to examine heterogeneity among the es- timates, controlling for additional 68 variables reflecting the design of primary studies. Based on the heterogeneity results, we find that the publication bias in our dataset is driven by 'lower quality' studies based on the RePEc ranking of economic journals. This finding also aligns with our robustness checks on sub- sets divided according to the quality of journals publishing the primary studies.
|
guest ::
