The Gallup Poll and General Social Survey have asked Americans about their religious service attendance since 1939 and 1972, respectively. With remarkable consistency, these two surveys have estimated that just over 40% of the American population regularly attends religious services. Yet, recent research has called this “gold standard” into question, citing three sources of bias in these estimates: (a) ambiguous item wording, (b) an ambiguously specified time frame; and (c) data collection methods that lend themselves to socially desirable responding. Several lines of research have developed to eliminate or minimize these sources of bias, but these efforts have yielded a wide variety of results, with some estimates being half as much as the gold standard! Methodological and psychometric differences are not the only source of variation, however. The characteristics of those sampled into studies also introduces variability. Given that attendance estimates are likely influenced by variations in both methodology and sampling, this study uses meta-analytic techniques to estimate the extent of their influence and to estimate the attendance rate after controlling for their influence. The findings indicate that efforts to reduce socially desirable responding have had the greatest impact on the attendance rate, followed by efforts to overcome the ambiguously specified time-frame. In addition, attendance rates are positively related to the proportion of African Americans, Whites and married respondents sampled, as well as mean years of education. Attendance rates are also negatively related to the proportion of 18 to 30 year-old respondents sampled. After controlling for these methodological and socio-demographic study characteristics, the prevalence of weekly attendance in America was variously estimated as 41.4% for the gold standard items, 43.1% for items measuring attendance in the past week, 27.8% when asking respondents what they did yesterday (i.e., on Sunday via the time-use methodology) and 22.7% when attendance was counted manually.

The elasticity of factor substitution between capital and labor is a crucial parameter in many economic fields. However, despite extensive research, there is no agreement on its value. Utilizing 738 estimates from 41 studies published between 1961 and 2016, this paper provides the first meta-regression analysis of capital-labor substitution elasticities for the U.S. economy. We show that heterogeneity in reported estimates is driven by the choice of estimation equations, the modeling of technological dynamics, and data characteristics. Based on the underlying meta-regression sample and a "best practice" specification, we estimate a long-run elasticity in the range of 0.6 to 0.7. For all estimated elasticities the hypothesis of a Cobb-Douglas production function is rejected.

Purchasing Power Party (PPP) has long been intensively studied in empirical researches. However, a unanimous conclusion has not been reached. As an alternative to a narrative literature review, this paper conducts a meta-regression analysis of a collection of thirty-three studies, in order to uncover the sources of variation in the empirical findings relating to PPP. We also test the validities of suggestions made by the narrative literature reviewers that the use of more years of data, more countries, more powerful tests, more general model specifications, and an allowance for non-linearity might mitigate the issue of PPP puzzle. We find that the proposition is true and that whether PPP holds in the long run mainly depends on the methodology employed, the regimes the data are sampled from, and the length of the sample of data that is used. When addressing the persistence of the deviations from PPP, it mostly depends on the methodologies adopted.

Objectives: The aim of this dissertation was to examine bias in randomized controlled trials (RCTs) and non-randomized studies (NRS) in surgery using the literature evaluating laparoscopy and conventional (i.e. open) surgery for the treatment of colon cancer as a case study. The objectives were 1) to develop a conceptual framework for bias in comparative NRS; 2) to compare effect estimates from NRS with those from RCTs at low risk of bias; 3) to explore the impact of NRS-design attributes on estimates of treatment effect. Methods: The methods included a modified framework synthesis, systematic review of the literature, random-effects meta-analyses, and frequentist and Bayesian meta-regression. The Cochrane Risk of Bias Tool was used to classify trials as Strong RCTs (i.e. low risk of bias) or Typical RCTs (i.e. unclear or high risk of bias). Results: A conceptual framework for bias in comparative NRS was developed and it contains 37 individual sources of bias or “items”. These items were organized within 6 overarching “domains”: selection bias, information bias, performance bias, detection bias, attrition bias, and selective reporting bias. Our analyses revealed that NRS were associated with more extreme estimates of benefit for laparoscopy than Strong RCTs when examining subjective outcomes. The odds ratios from NRS were 36% smaller (i.e. demonstrating more benefit for laparoscopy) than those from Strong RCTs for the outcome post-operative complications (Ratio of Odds Ratios, ROR 0.64, [0.42, 0.97], p=0.04). Similar exaggerated benefit was seen among NRS when assessing length of stay, (Difference in Mean Differences, -2.15 days, [-4.08, -0.21], p=0.03). This pattern was not observed with the objective outcomes peri-operative mortality and number of lymph nodes harvested. Analyses adjusted for period effects and between-study case-mix yielded similar findings. Finally, effect estimates in NRS did not consistently vary according to the presence or absence of nine design characteristics identified from the conceptual framework. Conclusions: We have demonstrated that the results of surgical NRS can be significantly biased as compared with those of low risk of bias RCTs when evaluating subjective outcomes. However, none of the nine NRS-design characteristics examined was consistently associated with biased effect estimates.

Objectives: The aim of this dissertation was to examine bias in randomized controlled trials (RCTs) and non-randomized studies (NRS) in surgery using the literature evaluating laparoscopy and conventional (i.e. open) surgery for the treatment of colon cancer as a case study. The objectives were 1) to develop a conceptual framework for bias in comparative NRS; 2) to compare effect estimates from NRS with those from RCTs at low risk of bias; 3) to explore the impact of NRS-design attributes on estimates of treatment effect. Methods: The methods included a modified framework synthesis, systematic review of the literature, random-effects meta-analyses, and frequentist and Bayesian meta-regression. The Cochrane Risk of Bias Tool was used to classify trials as Strong RCTs (i.e. low risk of bias) or Typical RCTs (i.e. unclear or high risk of bias). Results: A conceptual framework for bias in comparative NRS was developed and it contains 37 individual sources of bias or “items”. These items were organized within 6 overarching “domains”: selection bias, information bias, performance bias, detection bias, attrition bias, and selective reporting bias. Our analyses revealed that NRS were associated with more extreme estimates of benefit for laparoscopy than Strong RCTs when examining subjective outcomes. The odds ratios from NRS were 36% smaller (i.e. demonstrating more benefit for laparoscopy) than those from Strong RCTs for the outcome post-operative complications (Ratio of Odds Ratios, ROR 0.64, [0.42, 0.97], p=0.04). Similar exaggerated benefit was seen among NRS when assessing length of stay, (Difference in Mean Differences, -2.15 days, [-4.08, -0.21], p=0.03). This pattern was not observed with the objective outcomes peri-operative mortality and number of lymph nodes harvested. Analyses adjusted for period effects and between-study case-mix yielded similar findings. Finally, effect estimates in NRS did not consistently vary according to the presence or absence of nine design characteristics identified from the conceptual framework. Conclusions: We have demonstrated that the results of surgical NRS can be significantly biased as compared with those of low risk of bias RCTs when evaluating subjective outcomes. However, none of the nine NRS-design characteristics examined was consistently associated with biased effect estimates.

The primary aim of this thesis is to examine whether the policy recommendations made by the European Central Bank in response to the financial crisis of 2008 were biased towards fiscal consolidation. It posits that such policies, commonly known as austerity, were underpinned by estimates of the fiscal multiplier that were lower than those of international and independent researchers. To analyse this, it provides a systematic overview of the ECB's fiscal multiplier estimates by performing a meta-regression analysis on all ECB working papers making multiplier estimates published between 1992 and 2012, and comparing the results against those of a larger dataset containing multiplier estimates made. It finds that the multiplier estimates of the ECB are significantly lower than the norm, which is potentially suggestive of bias. This thesis contributes to the literature on ideational bias in economic policy-making by providing a systematic literature review that helps inform the discussion on austerity in the EU. It also servers as a replication and expansion of previous meta-regression studies on the fiscal multiplier, by being the first study that specifically examines the estimates of a specific institution.

Cette thèse s’intéresse aux effets des transferts de fonds sur la pauvreté et les différents indicateurs de bien-être et à la mesure de l’impact de ces flux monétaires sur les différentes positions de la répartition de la dépense par tête. L’étude utilise les données des deux plus récentes enquêtes sur les ménages du Vietnam (VHLSS 2010 et 2012). Plusieurs méthodes empiriques sont utilisées dans cette thèse: méta-analyse, appariement par propension de score, la régression avec les variables instrumentales et enfin l’effet du traitement sur le quantile (QTE). On a utilisé l’indicateur de pauvreté (basé sur la consommation) selon le critère du Bureau General des Statistiques du Vietnam (GSO) et de la Banque Mondiale (WB) et les indicateurs de bien-être : dépenses par tête et dépenses des ménages. Plusieurs résultats intéressants ont émergé. L'étude présentée dans le chapitre 2 constitue une des premières méta-analyses pour mettre en évidence l'effet empirique véritable de l'impact des transferts de fonds sur la pauvreté. Nos résultats ont trouvé que de nombreuses caractéristiques de l’étude peuvent expliquer l’hétérogénéité de différentes tailles reportées de l’effet des transferts de fonds sur la pauvreté : la région géographique et le niveau de développement de pays récipiendaire, la méthode économétrique et les indicateurs de pauvreté utilisés. Dans le chapitre 3, on a utilisé la méthode d’appariement par propension de score. On a trouvé que les transferts ont réduit la probabilité d’être pauvre entre 7.3 % pour les familles rurales et 3 % pour les familles urbaines. En outre, les transferts de fonds ont amélioré le niveau de vie des ménages qui en bénéficient. Dans le chapitre 4, nous avons utilisé la méthode de régression avec les variables instrumentales, les transferts de fonds ont réduit la probabilité pour la famille de vivre sous le seuil de pauvreté de 7.5%. Les résultats sont robustes et vérifiés par les différentes méthodes économétriques. Au final, on trouve que le gain de transfert de fonds est plus important dans les familles riches au Vietnam, surtout quand on prend en compte l’endogenéité de la variable «transfert de fonds ». / This study examines the impact of remittances on poverty, the welfare of receiving-households. It measures the extent to which the remittance flows are distributed among households at various points in the consumption distribution, using two latest rounds of Vietnam Household Living Standard Surveys 2010 and 2012 (VHLSSs). We applied different estimation method in this study including the meta-regression framework, matching method, IV method and quantile treatment effect regression. We use the expenditure poverty line measured by General Statistics Office/World Bank (GSO-WB), while the welfare indicator is measured by household and per capita expenditure. Several interesting results emerged. In chapter 2, the study is one of the first meta-regression analysis regarding the investigation of the “true” empirical effect of remittances on poverty reduction. The findings reveal that the heterogeneity in effect size of the impact of remittance on poverty reduction for a specific study may depend on various key factors: geographical region and level of development of one country, econometric method, and poverty indicators. In chapter 3, we apply the matching method. The poverty reduction effect of remittances varies between 7.3 percentage points and 3.0 percentage points for rural and urban families respectively. Furthermore, we find a significant increase of expenditure for remittances-receiving households. In chapter 4, using the historical migration networks as instruments for remittance receipts, we found that international remittances do decrease the likelihood of household poverty by about 7.5 percentage points. Finally, the gain of remittances is significantly larger among better–off households in Vietnam, especially when the endogeneity of remittance is taken into account in QTE estimation.

Introduction: There has been increasing interest in evidence from pragmatic trials as healthcare providers and decision makers must determine if available evidence can be translated and used in real world practice. As a result, a number of tools have been developed to help researchers design and appraise randomized controlled trials (RCTs) within the pragmatic-explanatory continuum. It is unclear what role pragmatism plays in heterogeneity and if pragmatic and explanatory trials should be pooled in meta-analyses of systematic reviews. Objectives: Our primary objective was to explore the role of pragmatism (based on the Pragmatic-Explanatory Continuum Indicator Summary-2 [PRECIS-2] score) as a source of heterogeneity in Cochrane systematic reviews with at least substantial heterogeneity (I2≥ 50%). Our secondary objective was to compare and contrast the application of the established PRECIS-2 tool to the newly developed Rating of Included Trials on the Efficacy-Effectiveness Spectrum (RITES) tool. Methods: We conducted a cross-sectional methodological review on systematic reviews of RCTs published in the Cochrane Library from January 1, 2014 to January 1, 2017. Included systematic reviews had a minimum of 10 RCTs in the meta-analysis of the primary outcome and at least moderate heterogeneity (I2≥ 50%). Of the eligible systematic reviews, a random selection of 10 were included for quantitative evaluation. In each systematic review, RCTs were scored using the PRECIS-2 and RITES tools, in duplicate, to determine the amount of pragmatism. Meta-regression modelling was performed to evaluate how much variability in heterogeneity (quantified by I2) was due to pragmatism. Inter-rater reliability of both PRECIS-2 and RITES was measured using the intraclass correlation coefficient and Spearman’s correlation coefficient was used to determine the strength of the relation between PRECIS-2 and RITES. Results: Ten systematic reviews from nine Cochrane Review Groups were included in the quantitative analysis. The reviews included an average of 13 RCTs (standard deviation=2.6) for a total of 132 RCTs of which 128 could be obtained. When the PRECIS-2 summary score was entered as a covariate in random effects meta-regression models for each systematic review, there were minimal changes in heterogeneity. The changes in I2 ranged from 0.2% to 13.3%. Conclusion: Based on these findings it appears pragmatism as measured by PRECIS-2 does not explain heterogeneity in systematic reviews, therefore pooling of pragmatic and explanatory RCTs is unlikely to be detrimental to meta-analyses. / Thesis / Master of Science (MSc) / Systematic reviews and meta-analyses of randomized controlled trials (RCTs) are an important scientific activity that can lead to changes in health care. However, there is concern whether it is appropriate to meta-analyze data from RCTs that are performed under more controlled conditions (explanatory RCTs) and RCTs that are performed under more real world conditions (pragmatic RCTs) since there may be variability between them. The purpose of this research was to explore how much these trial types affect variability, otherwise known as heterogeneity, in systematic reviews. We applied a scoring tool called the Pragmatic-Explanatory Continuum Indicator Summary-2 (PRECIS-2) to RCTs within 10 systematic reviews with at least moderate heterogeneity and performed statistical modelling to determine how much heterogeneity could be explained by a trial being more or less pragmatic. Results showed that trial type did not explain heterogeneity therefore it is probably reasonable to meta-analyze data from pragmatic and explanatory RCTs.

A Quantitative Microbial Risk Assessment (QMRA) has been undertaken to utilize research on Shiga-toxin Escherichia coli (STEC) contamination in beef for the benefit of public health. The QMRA operates as a 2nd order Monte Carlo simulation to create stochastic mathematical models that incorporate all of the key components of STEC contamination from farm to fork. The resulting model is able to identify knowledge gaps, public health risks, and simulate theoretical changes in the beef system. However, high variability in processing plant intervention literature has prompted a meta-analysis to determine informed estimates of intervention effectiveness for QMRA parameterization. Meta-analysis derived least-squares means bacterial log reductions for acetic acid, lactic acid, steam vacuum, and water wash interventions on carcass surfaces (n=249) were 1.44 [95% CI: 0.73 – 2.15], 2.07 [1.48 – 2.65], 3.09 [2.46 – 3.73], and 1.90 [1.33 – 2.47] log CFU/cm2, respectively. Least-squares means log reductions for acetic acid, lactic acid, sodium hydroxide, and water wash on hide surfaces (n=47) were 2.21 [1.36 – 3.05], 3.02 [2.16 – 3.88], 3.66 [2.60 – 4.72], and 0.08 [-0.94 – 1.11] log CFU/cm2, respectively. Meta-regressions showed that temperature, duration of application, microbial starting concentration, extra water washes, inoculation type, organism type, sample method, surface type, and antimicrobial concentrations were all significant predictors of intervention effectiveness. Finally, after observing authors use substituted values for samples found below a detection limit in primary plant intervention literature, simulations were run to assess the impact of substitution on a random-effects meta-analysis. Simulation results show that substitution practices artificially decrease effectiveness estimates and increase heterogeneity. / Master of Science

The elasticity of factor substitution between capital and labor is a crucial parameter in many economic fields. However, despite extensive research, there is no agreement on its value. Utilizing 738 estimates from 41 studies published between 1961 and 2016, this paper provides the first meta-regression analysis of capital-labor substitution elasticities for the U.S. economy. We show that heterogeneity in reported estimates is driven by the choice of estimation equations, the modeling of technological dynamics, and data characteristics. Based on the underlying meta-regression sample and a "best practice" specification, we estimate a long-run elasticity in the range of 0.6 to 0.7. For all estimated elasticities the hypothesis of a Cobb-Douglas production function is rejected.