Stata软件在元分析异质性检验中的运用

研究间异质性检验是合并各独立研究效应量的前提条件.利用Stata软件不仅可制作森林图和Galbraith星状图进行定性分析,而且可以计算Q、I2和H统计量进行定量分析.综合使用两种分析方法能够实现对异质性的准确判断.     

XRCC1 Arg399Gln and clinical outcome of platinum-based treatment for advanced non-small cell lung cancer: a meta-analysis in 17 studies

Objective: XRCC1 polymorphism is a research hotpot in individual treatment for non-small cell lung cancer (NSCLC). To obtain the association between XRCC1 polymorphism and clinical outcome of platinum-based treatment for NSCLC, a meta-analysis was conducted. Methods: Databases including PubMed, Embase, Cochrane, and Chinese National Knowledge Infrastructure (CNKI) were searched for publications that met the inclusion criteria. A fixed effect model was used to estimate pooled odds ratio (OR) and hazard ratio (HR) with 95% confidence interval (CI) for the association between XRCC1 Arg399Gln and response or survival of platinum-based treatment for advanced NSCLC. A chi-squared-based Q-test was used to test the heterogeneity hypothesis. Egger’s test was used to check publication bias. Results: Seventeen published case-control studies that focus on the association between XRCC1 Arg399Gln and response or survival of platinum-based treatment for advanced NSCLC in 2 256 subjects were included in this meta-analysis, of whom 522 were AA genotypes (23.2% frequency), 916 AG genotypes (40.6% frequency), and 818 GG genotypes (36.2% frequency). The overall response rate (ORR) was 45.2% (110/243) for AA genotype patients, 29.9% for AG genotype (73/244), and 30.7% for GG genotype (124/403). The heterogeneity test did not show any heterogeneity and the Egger’s test did not reveal an obvious publication bias among the included studies. The meta-analysis indicated that AA genotype patients presented higher response rates toward platinum drug treatment compared with G model (GG+GA) patients (GG vs. AA model: OR=0.489, 95% CI 0.266–0.900, P=0.021; AG vs. AA model: OR=0.608, 95% CI 0.392–0.941, P=0.026; GA+AA vs. GG model: OR=1.259, 95% CI 0.931–1.701, P=0.135; GG+GA vs. AA model: OR=0.455, 95% CI 0.313–0.663, P=0.0001). However, no evidence validates XRCC1 associates with the survival following platinum drug therapy. Conclusions: Our meta-analysis suggested that XRCC1 Arg399Gln is related with the sensitivity of NSCLC patients to platinum-based treatment. AA genotype patients present more desirable curative effectiveness compared with other patients.     

Bias Corrections for Standardized Effect Size Estimates Used With Single-Subject Experimental Designs

A multilevel meta-analysis can combine the results of several single-subject experimental design studies. However, the estimated effects are biased if the effect sizes are standardized and the number of measurement occasions is small. In this study, the authors investigated 4 approaches to correct for this bias. First, the standardized effect sizes are adjusted using Hedges’ small sample bias correction. Next, the within-subject standard deviation is estimated by a 2-level model per study or by using a regression model with the subjects identified using dummy predictor variables. The effect sizes are corrected using an iterative raw data parametric bootstrap procedure. The results indicate that the first and last approach succeed in reducing the bias of the fixed effects estimates. Given the difference in complexity, we recommend the first approach.     

A meta-analysis of instructional systems applied in science teaching

This article is a report of a meta-analysis on the question: “What are the effects of different instructional systems used in science teaching?” The studies utilized in this meta-analysis were identified by a process that included a systematic screening of all dissertations completed in the field of science education since 1950, an ERIC search of the literature, a systematic screening of selected research journals, and the standard procedure of identifying potentially relevant studies through examination of the bibliographies of the studies reviewed. In all, the 130 studies coded gave rise to 341 effect sizes. The mean effect size produced over all systems was 0.10 with a standard deviation of 0.41, indicating that, on the average, an innovative teaching system in this sample produced one-tenth of a standard deviation better performance than traditional science teaching. Particular kinds of teaching systems, however, produced results that varied from this overall result. Mean effect sizes were also computed by year of publication, form of publication, grade level, and subject matter.     

PowerUp!: A Tool for Calculating Minimum Detectable Effect Sizes and Minimum Required Sample Sizes for Experimental and Quasi-Experimental Design Studies

Abstract

This paper and the accompanying tool are intended to complement existing supports for conducting power analysis tools by offering a tool based on the framework of Minimum Detectable Effect Sizes (MDES) formulae that can be used in determining sample size requirements and in estimating minimum detectable effect sizes for a range of individual- and group-random assignment design studies and for common quasi-experimental design studies. The paper and accompanying tool cover computation of minimum detectable effect sizes under the following study designs: individual random assignment designs, hierarchical random assignment designs (2-4 levels), block random assignment designs (2-4 levels), regression discontinuity designs (6 types), and short interrupted time-series designs. In each case, the discussion and accompanying tool consider the key factors associated with statistical power and minimum detectable effect sizes, including the level at which treatment occurs and the statistical models (e.g., fixed effect and random effect) used in the analysis. The tool also includes a module that estimates for one and two level random assignment design studies the minimum sample sizes required in order for studies to attain user-defined minimum detectable effect sizes.     

The development of students critical thinking abilities and dispositions through the concept mapping learning method – A meta-analysis

Our meta-analysis investigates the effectiveness of the concept mapping method compared with the traditional teaching method, for the improvement of critical thinking ability and disposition outcomes. We included studies measuring students' critical thinking through standardised tests, comparing an experimental (concept mapping) group with a control (traditional teaching) one. Effect sizes for critical thinking ability and disposition outcomes were pooled with a random-effects model. We included in our meta-analysis 21 studies (108 comparisons) involving 1695 students. The concept mapping method was more effective than the traditional teaching method at posttest for the improvement of critical thinking ability (g = 0.531, 95% CI 0.279 to 0.783) and disposition (g = 0.648, 95% CI 0.266 to 1.031). Heterogeneity was moderate to high, Egger's test did not indicate any evidence of publication bias; however, both visual inspection of the Funnel plot and Duval and Tweedie's trim and fill method indicated potentially three missing studies for critical thinking ability and two for critical thinking disposition. Additionally, we analysed the potential moderating effect of students' demographic characteristics, educational conditions, concept mapping elaboration methods and allocation type. Allocation type was a significant moderator, having a strong effect on concept mapping for critical thinking abilities in randomised studies (g = 0.739, 95% CI 0.356 to 1.122), but its effect is low in non-randomised studies (g = 0.265, 95% CI 0.014 to 0.517).     

A meta-analysis of the effect of education on social capital

To assess the empirical estimates of the effect of education on social trust and social participation – the basic dimensions of individual social capital – a meta-analysis is applied, synthesizing 154 evaluations on social trust, and 286 evaluations on social participation. The publication bias problem is given special emphasis in the meta-analysis. Our statistical synthesis confirms that education is a strong and robust correlate of individual social capital. The meta-analysis provides support for the existence of a relative effect of education on social participation, and of a reciprocity mechanism between the dimensions of social capital. The analysis also suggests that the erosion of social participation during the past decades has coincided with a decrease of the marginal return to education on social capital. Finally, we find differences in the return to education between genders, between US and other nations, and variations for different education attainments.     

What limits the encoding effect of note-taking? A meta-analytic examination

Previous meta-analyses indicate that the overall encoding effect of note-taking is positive but modest. This meta-analysis of 57 note-taking versus no note-taking comparison studies explored what limits the encoding effect by examining the moderating influence of seven variables: intervention, schooling level, presentation mode and length, test mode, and publication year and source. It was found that (a) either positive interventions or rise in schooling level did not enhance the benefits of note-taking; (b) visual presentation of learning material interfered with the note-taking process, whereas longer presentation did not; (c) recall test detected the encoding effect more than recognition and higher-order performance tests; and (d) publication year and source contributed to the variation in effect sizes. These results suggest that the modest encoding effect is not due to the incompleteness of students’ spontaneous note-taking procedures, but mechanical demands of note-taking, type of learning outcome measure, and publication characteristics.     

教育对中国经济增长的影响效应——基于元回归分析方法的分析

基于60篇教育与中国经济增长实证研究中626个回归方程的元分析发现,教育对中国经济增长的作用估计受到不同研究特征的影响。尤其是,"显著效应"的实证结果受到估计方法、教育变量和经济增长变量的设定形式、教育变量的测量指标、样本起始年份、区域和控制变量选择等研究特征的显著影响。使用变量增长率形式、运用工具变量法、使用入学率和教育经费数据估计会降低显著性效应的概率;样本区域覆盖面越小、样本起始年份越早、使用受教育年限数据估计会增加显著性效应的概率。漏斗不对称检验显示样本文献存在显著的正向偏误,即回归结果越显著的研究越有可能被发表。因此,在分析教育对我国经济增长的作用时,必须考虑研究之间的异质性和可能存在的发表偏误。未来应重点关注的是,改进测量和估计方法,通过更严谨的研究设计精确识别教育对我国经济增长的因果效应;结合中国经济发展的阶段特征,分行业和部门讨论教育的作用,同时拓展非线性模型以考察教育的收益递增的外部效应;持续关注对教育质量的监测及其对经济增长的影响,为高质量教育体系的建设提供科学的理论支持。     

Parental involvement on student academic achievement: A meta-analysis

This paper is a quantitative synthesis of research into parental involvement and academic achievement through a meta-analysis of 37 studies in kindergarten, primary and secondary schools carried out between 2000 and 2013. Effect size estimations were obtained by transforming Fisher's correlation coefficient. An analysis has also been conducted of the heterogeneity of the magnitudes grouped according to different moderator variables, and a study of the publication bias affecting meta-analytical studies. The results show that the parental models most linked to high achievement are those focusing on general supervision of the children's learning activities. The strongest associations are found when the families have high academic expectations for their children, develop and maintain communication with them about school activities, and help them to develop reading habits.     

Modeling the impacts of child care quality on children's preschool cognitive development

The National Institute of Child Health and Human Development (NICHD) Study of Early Child Care compared 3 statistical methods that adjust for family selection bias to test whether child care type and quality relate to cognitive and academic skills. The methods included: multiple regression models of 54-month outcomes, change models of differences in 24- and 54-month outcomes, and residualized change models of 54-month outcomes adjusting for the 24-month outcome. The study was unable to establish empirically which model best adjusted for selection and omitted-variable bias. Nevertheless, results suggested that child care quality predicted cognitive outcomes at 54 months, with effect sizes of .04 to .08 for both infant and preschool ages. Center care during preschool years also predicted outcomes across all models.     

Evaluating the evidence base for evidence-based instructional practices in chemistry through meta-analysis

Investigating the effectiveness of instructional practices provides an evidence base to inform instructional decisions. Synthesizing research studies on instructional effectiveness provides an estimate of the generalizability of effectiveness across settings, along with an exploration of factors that may moderate the impact, which cannot be achieved within individual studies. This study sought to provide a synthesis of evidence-based instructional practices (EBIPs) particular to chemistry through meta-analysis. Ninety-nine studies were analyzed comprising a broader view of chemistry specific studies than past meta-analyses. The results showed that EBIPs feature a demonstrably positive impact on students' academic performance in chemistry, although assessment topic coverage and setting size emerged as relevant moderators of impact and prevented making definitive conclusions of the relative impact of each EBIP. In examining publication bias, an asymmetric distribution of studies based on standard error (SE) and effect size was found, indicative of potential publication bias. To explore the potential impact of bias, the trim and fill method was employed resulting in a range for the overall weighted effect size from 0.29 to 0.62. The study concludes that evidence-based instructional practices have demonstrated effectiveness even in consideration of potential publication bias, as the range of effect sizes remains positive, but highlights the continued need to publish null findings in the research literature.     

Bias Reduction in Quasi-Experiments With Little Selection Theory but Many Covariates

Abstract: In observational studies, selection bias will be completely removed only if the selection mechanism is ignorable, namely, all confounders of treatment selection and potential outcomes are reliably measured. Ideally, well-grounded substantive theories about the selection process and outcome-generating model are used to generate the sample of covariates. However, covariate selection is more heuristic in actual practice. Using two empirical data sets in a simulation study, we investigate four research questions about bias reduction when the selection mechanism is not known but many covariates are measured: (1) How important is the conceptual heterogeneity of the covariate domains in the data set? (2) How important is the number of covariates assessing each domain? (3) What are the joint effects of this conceptual heterogeneity and of the number of covariates per domain? (4) What happens to bias reduction when the set of covariates is deliberately impoverished by removing the covariates most responsible for selection bias, thus ensuring a slightly smaller but still heterogeneous set of covariates? The results indicate: (1) increasingly more bias is reduced as the number of covariate domains and the number of covariates per domain increase, though the rate of bias reduction is diminishing in each case; (2) sampling covariates from multiple heterogeneous covariate domains is more important than choosing many measures from fewer domains; (3) the most heterogeneous set of covariate domains removes almost all of the selection bias when at least five covariates are assessed in each domain; and (4) omitting the most crucial covariates generally replicates the pattern of results due to the number of domains and the number of covariates per domain, but the amount of bias reduction is less than when all variables are included and will surely not satisfy all consumers of causal research.     

Meta-analysis of the effect of practical reasoning instruction on student outcome in Home Economics education in Korea

This meta-analysis reviewed the effect of practical reasoning instruction (PRI) on student outcome in Home Economics education in Korea. In this meta-analysis, 25 studies with 35 effect sizes were analyzed. The results of this review indicated that PRI is more effective than traditional instruction on student outcomes. A medium and significant mean effect was 0.60 (SE?=?0.06). Categorical analyses and regression analyses were employed to find the sources of variance and moderators that predict the effects of PRI. The moderator analyses revealed no statistically significant effects of publication status, study design, type of student outcome, gender of students, and location. The school level of the students and duration were revealed to be moderators. The effect of PRI on student outcome was found to be smaller in middle schools than in elementary or high schools. The results of regression analysis for middle school students indicated that the effect of PRI was predicted by the length of the intervention. A sensitivity analysis indicates that the method this review used for calculating effect sizes for nonexistent control groups was robust.     

Do instructional interventions influence college students’ critical thinking skills? A meta-analysis

Promoting students’ critical thinking skills is an important task of higher education. Colleges and universities have designed various instructional interventions to enhance students’ critical thinking skills. Empirical studies have yielded inconsistent results in terms of the effects of such interventions. This meta-analysis presents a synthesis of empirical studies designed to promote measurable changes in students’ critical thinking skills using instructional interventions. Findings demonstrated statistically significant but small average effect size and evidence of heterogeneity among studies. Hierarchical linear model was adopted to explore potential predictors of the variance across effect sizes. Results showed that student discipline and treatment length explained part of the variability among treatment effects. Limitations and implications are discussed.     

A Meta-Analysis on Gender Differences in Mental Rotation Ability Measured by the Purdue Spatial Visualization Tests: Visualization of Rotations (PSVT:R)

This meta-analysis was conducted to estimate the magnitude of gender difference in three-dimensional (3-D) mental rotation ability and to investigate how factors related to test administration conditions play a role in varying gender difference effect sizes and threatening validity. Individuals’ 3-D mental rotation ability was measured by the Purdue Spatial Visualization Tests: Visualization of Rotations (PSVT:R). We integrated 70 effect sizes of gender differences in mental rotation ability measured by the PSVT:R which were obtained from 40 primary studies. The results indicated that male participants outperformed females on the test (Hedges’ g?=?0.57). The I ² statistic indicated 41.7 % of variation in effect sizes reflects real heterogeneity. The moderator analysis indicated that male superiority on spatial ability tasks measured by the PSVT:R is related to the implementation of time limits. The gender difference became larger when stringent time limits (equal or less than 30 s per item) were implemented.     

To Math or Not to Math: The Algebra-Calculus Pipeline and Postsecondary Mathematics Remediation

This article reports on a study designed to estimate the effect of high school coursetaking in the algebra-calculus pipeline on the likelihood of placing out of postsecondary remedial mathematics. A nonparametric variant of propensity score analysis was used on a nationally representative data set to remove selection bias and test for an effect among all comparable students and among students least likely to enroll in precalculus, trigonometry, or calculus. The effect failed to reach practical significance among all comparable students and was particularly weak among students least likely to take precalculus, trigonometry, or calculus. These findings suggest that the regression-based results that factor strongly into discussions on high school graduation requirements for mathematics coursetaking contain substantial selection bias.