Charles Perkins Centre, University of Sydney, Sydney, New South Wales, Australia.
School of Life and Environmental Sciences, University of Sydney, Sydney, New South Wales, Australia.
Res Synth Methods. 2020 Jul;11(4):553-567. doi: 10.1002/jrsm.1423. Epub 2020 Jun 15.
Meta-analyses are often used to estimate the relative average values of a quantitative outcome in two groups (eg, control and experimental groups). However, they may also examine the relative variability (variance) of those groups. For such comparisons, two relatively new effect size statistics, the log-transformed "variability ratio" (the ratio of two standard deviations; lnVR) and the log-transformed "coefficients of variation ratio" (the ratio of two coefficients of variation; lnCVR) are useful. In practice, lnCVR may be of most use because a treatment may affect the mean and the variance simultaneously. We propose new estimators for lnCVR and lnVR, including for when the two groups are dependent (eg, cross-over and pre-test-post-test designs). Through simulation, we evaluated the bias of these estimators and make recommendations accordingly. We use the methods to demonstrate that: (a) lifestyle interventions have a heterogenizing effect on gestational weight gain in obese women and (b) low-glycemic index (GI) diets have a homogenizing effect on glycemic control in diabetics. We also find that the degree to which dependence among samples is accounted for can impact parameters such as τ (ie, the between-study variance) and I (ie, the proportion of the total variability due to between-study variance), and even the overall effect, and associated qualitative interpretations. Meta-analytic comparison of the variability between two groups enables us to ask completely new questions and to gain fresh insights from existing datasets. We encourage researchers to take advantage of these convenient new effect size measures for the meta-analysis of variation.
荟萃分析常用于估计两组（例如对照组和实验组）之间定量结果的相对平均值。但是，它们也可以检查这些组的相对变异性（方差）。对于这种比较，两种相对较新的效应量统计数据，即对数转换的“变异比”（两个标准差之比；lnVR）和对数转换的“变异系数比”（两个变异系数之比；lnCVR）很有用。在实践中，lnCVR 可能最有用，因为处理可能会同时影响均值和方差。我们提出了 lnCVR 和 lnVR 的新估计量，包括当两组相关时（例如交叉设计和预测试后测试设计）。通过模拟，我们评估了这些估计量的偏差，并相应地提出了建议。我们使用这些方法来证明：（a）生活方式干预对肥胖女性的妊娠期体重增加有异质化作用，（b）低血糖指数（GI）饮食对糖尿病患者的血糖控制有同质化作用。我们还发现，样本之间的依赖性程度会影响参数，例如 τ（即研究之间的方差）和 I（即由于研究之间的方差导致的总变异性的比例），甚至影响整体效果和相关的定性解释。对两组之间的变异性进行荟萃分析使我们能够提出全新的问题，并从现有数据集获得新的见解。我们鼓励研究人员利用这些方便的新效应量措施来进行变异的荟萃分析。
