Measurement of associations in periodontal diseases using statistical methods for dependent data.
Measurement of associations in periodontal diseases using statistical methods for dependent data.
复制标题
使用相关数据的统计方法测量牙周疾病的关联性。
DOI：
10.1111/j.1600-0765.1991.tb01648.x
复制
发表时间：
1991
影响因子：
3.5
通讯作者：
Hujoel,P
中科院分区：
文献类型：
--
作者：
DeRouen,TA;Mancl,L;Hujoel,P
牙周病的临床研究通常涉及在每个研究患者的多个部位获得测量变量的观察结果。因此，这些观察结果本质上是相互依赖的，使用标准统计方法来衡量变量之间的关联是不合适的，因为它们通常假设观察结果之间是独立的。方法建议扩展标准的统计方法的情况下，数据是依赖的。例如，如果对诊断测试的灵敏度和特异性感兴趣，则假设独立性的通常二项分布可以用相关二项分布代替，并且在存在相关数据的情况下仍然可以估计这些特征。在通常使用多元线性或逻辑回归研究几个变量与反应的相互关系的情况下，相关数据的存在，例如来自牙周病临床研究的数据，需要进行调整以说明相关性。考虑到这种相关性的最有前途的方法之一是使用广义估计方程（GEE）的方法，这是本文所描述的，并在两个例子中。这两个例子涉及测量天冬氨酸氨基转移酶（AST）水平与组织破坏的相关性，以及评价甲硝唑和手术对不同慢性牙周炎患者组中附着水平的影响。几种统计方法和他们的结论进行比较，为每个例子。GEE方法允许将标准回归模型扩展到涉及相关数据的情况，同时保持与回归结果相关的解释的容易性。
Clinical studies of periodontal diseases often involve obtaining observations for variables measured at multiple sites within each patient studied. These observations are thus inherently dependent, and the use of standard statistical methods to measure associations among the variables is inappropriate since they ordinarily assume independence among observations. Methods are suggested for extending standard statistical methods to the situation where data are dependent. For example, if the sensitivity and specificity of a diagnostic test is of interest, the usual binomial distribution, which assumes independence, can be replaced with the correlated binomial, and these characteristics can still be estimated in the presence of correlated data. In situations where the interrelationships of several variables to a response would ordinarily be investigated using multiple linear or logistic regression, the presence of correlated data such as that arising from clinical studies of periodontal disease requires that adjustments be made to account for the correlation. One of the most promising methods for taking such correlation into account is the use of generalized estimating equation (GEE) methodology, which is described in this paper and employed in two examples. The two examples involve measuring the association of levels of aspartate aminotransferase (AST) with tissue destruction, and the evaluation of the effects of metronidazole and surgery on attachment level in different groups of chronic periodontitis patients. Several statistical approaches and their conclusions are compared for each of the examples. GEE methodology allows the extension of standard regression models to situations involving dependent data while maintaining the ease of interpretation associated with regression results.
登录
查看更多内容
影响因子：
3.5
作者：
PERSSON, GR;DEROUEN, TA;PAGE, RC
通讯作者：
PAGE, RC
影响因子：
2
作者：
LOE, HARALD;SILNESS, JOHN
通讯作者：
SILNESS, JOHN
通过减影射线照相测量牙周治疗对牙槽骨的影响。
影响因子：
4.3
作者：
Schmidt,EF;Webber,RL;Ruttimann,UE;Loesche,WJ
通讯作者：
Loesche,WJ
影响因子：
1.9
作者：
ROSNER, B
通讯作者：
ROSNER, B
评估特定地点变量之间的关系。
DOI：
--
发表时间：
1990
期刊：
The Journal of Periodontology
影响因子：
--
作者：
P. Hujoel;W. Loesche;T. Derouen
通讯作者：
T. Derouen