第十章 临床科研中常见的偏倚及防治方法.ppt

* * * * * * * * * * * * * * * * * * * * * * * * * * * Right, in a case-reference study, the measure of association of choice is an odds ratio and this 2x2 table gives the framework for calculating it. So, is he on to something, this tobacco company epidemiologist? Does exposure to matches - and all those burnt carbons- cause lung cancer? Why not? * So, after stratifying by smoking status, you get 2 two-by-two tables. In each one, all subjects are homogeneous in terms of smoking. In this table, all the subjects are smokers and in this one all the subjects are non-smokers. What do we see? The effect of matches seen in the crude table is no longer seen in the smoking-specific strata. What we have formally done here is adjust for smoking and after adjustment, we no longer see an effect of matches in the development of lung cancer. * * * So, off went the nightlight and we began to stumble around her room. * These authors believed that the apparent association use of a night light and child’s myopia was likely because of confounding by parent’s myopia. These authors provided direct data that nearsightedness in parents is associated with the use of night light, probably to help the parents see better. And, there is data to suggest that myopia is inherited; ie if parents have it, their kids are more likely to have it. So, this is an example, where parenteral myopia confounds the relationship between night lights and child’s myopia and there is no real direct, or what we call independent, association between use of night lights and child’s myopia. * So, our fears relieved about night lights. . .. on comes the night light again and everyone is happy. * * * * * * * * * * * 读书 * * 疾病严重程度、患者经济状况、就诊方便与否都会影响入院率 * * * * * * * * * * * * * * * * * * * * * 随机分配方法可分为简单随机分组和分层随机分组。前者指按照随机分配原则，利 用随机化方法直接将研究对象分配到各组中。 (二)实施阶段 1．盲法分为单盲、双盲和三盲 单盲指只有研究者了解研究对象的分组情况而研究对象不知道自己是在试验组 还是在对照组。其优点是研究者可以更好地观察了解研究对象，可避免来自研究对 象方面的偏倚。缺点是避免不了研究者 方面带来的主观偏倚，易造成对试验组和对照组的处理不均衡。 双盲指研究者和研究对象都不了解试