Is it simple or simplistic?
Is it simple or simplistic?
复制标题
DOI：
10.1097/00005650-200305000-00007
复制
发表时间：
2003-05-01
期刊：
影响因子：
3
通讯作者：
Wyrwich, KW
中科院分区：
文献类型：
--
作者：
Norman, GR;Sloan, JA;Wyrwich, KW
对于一个普遍的观点，不可避免的反驳是它不是;对于一个简单的观点，它是复杂的;对于一个大多数值都在0.5左右的观点，有些值不是。讨论者正确地指出，我们的研究结果不能被认为是0.5 SD的普遍性。仍然必须仔细考虑条件，成本和治疗的好处等之间的个体差异，虽然我们同意，我们敦促认识到0.5 SD是一个指导方针，而不是一个规则，并不需要在所有情况下同样适用。我们和讨论者之间的分歧点似乎是玻璃杯是半满还是半空--0.5 SD应该被视为有用的近似值还是应该被抛弃。虽然后者可能更舒适，但它无助于推动生活质量评估科学的发展。讨论者似乎得出相反的结论，即0.5 SD指南要么太大，要么太小，突出了科学的变幻莫测。Wright表示，“……在没有风险的情况下，我相信几乎任何改变都会被病人认为是有益的。”这没有抓住重点;我们表明，变化必须高于可检测的歧视水平。Beaton引用了一个统计标准，等于1.96 SD（变化）（对于0.7的可靠性，约等于1.5的效应量），并得出结论，我们正在追求“阈值可能小于测量误差”。[1]但0.5标准差标准是一种观察，而不是一种观点。它既不太大，也不太小;它是观察到的，与关于歧视限度的心理学文献相一致。
The inevitable rejoinder to a claim that something is universal is that it’s not; to a claim that something is simple is that it’s complicated; to a claim that most values are around 0.5 is that some values are not. The discussants correctly point out that our findings cannot be taken as a claim to universality of 0.5 SD. Careful consideration still must be given to individual differences among conditions, costs, and benefits of therapies, etc. While we agree, we urge recognition that 0.5 SD is a guideline, not a rule, and need not apply equally in all circumstances. The point of disagreement between ourselves and the discussants would seem to be whether the glass is half full or half empty—should 0.5 SD be viewed as a useful approximation or should it be discarded. While the latter is potentially more comfortable, it does nothing to move the science of quality of life assessment forward.The vagaries of the science are highlighted in the discussants’ seemingly opposite conclusions, that the 0.5 SD guideline is either too big or too small. Wright states that “... in the absence of risks, I believe virtually any change would be perceived by patients as beneficial.” That misses the point; we showed that changes must be above the level of discrimination to be detectable. Beaton invokes a statistical criterion, equal to 1.96 SD (change)(which for a reliability of 0.7, is about equal to an effect size of 1.5) and concludes that we are pursuing “thresholds likely to be less than the measurement error.” 1 But the 0.5 SD criterion is an observation, not an opinion. It is neither too large nor too small; it is what was observed and is consistent with a psychological literature on the limit of discrimination.