WHAT WE NOW KNOW ABOUT I DONT KNOWS
WHAT WE NOW KNOW ABOUT I DONT KNOWS
复制标题
DOI：
10.1086/268217
复制
发表时间：
1975-01-01
影响因子：
3.4
通讯作者：
BUSCH, L
中科院分区：
文献类型：
--
作者：
FRANCIS, JD;BUSCH, L
调查研究人员面临的一个长期问题是如何处理“不知道”和“没有意见”的回答。对于“不回应”（拒绝和“不在家”）和“不适用”编码等类似问题，已经制定了普遍接受的准则。但不存在处理“不知道”和“无意见”答复的指南。造成这种情况的原因之一是对该问题很少进行系统的调查。相反，每个研究人员都必须对响应的性质和/或其分布做出单独的假设。本研究的重点是对无反应的随机性假设进行调查。下面的研究结果表明它是无效的。聚合分布仅提供支持（或反对）随机性假设的最弱证据。通过确定非实质性答复是否与受访者属性或数据收集程序的特征系统相关，可以获得更好的证据。此外，总体分布论证忽略了假设检验的影响点。调查分析师检验的假设很少涉及调查问卷中的全部项目。如果非实质性回答是系统性的，并且如果用于检验假设的项目集出现大量非实质性回答，那么如果处理不当，这些非实质性回答（NSR）将对测试产生某种系统性影响。
A perennial problem facing survey researchers is what to do with" don't know" and" no opinion" responses. For the kindred problems of" nonresponse"(refusals and" not at homes") and" not applicable" codings, commonly accepted guidelines have evolved. But no guidelines exist for handling" don't know" and" no opinion" responses. One of the reasons for this is that little systematic investigation of the problem has been made. Instead, each researcher has been left to make individual assumptions about the nature of the responses and/or their distribution. This study focuses on an investigation of the assumption of ran-domness in nonresponse. The research results presented below suggest that it is invalid. Aggregate distributions provide only the weakest evidence for (or against) the hypothesis of randomness. Better evidence is obtained by determining whether a nonsubstantive response is systematically related to respondent attributes or to characteristics of the data-gathering procedure. Also, the aggregate distribution argument misses the point of the effects of hypothesis testing. Hypotheses tested by survey analysts seldom involve the full set of items in their questionnaires. If nonsubstantive responses are systematic and if a large number appear for the set of items employed for testing the hypothesis, these non-substantive responses (NSRs) are going to have some kind of systematic effect on the test if they are treated inappropriately.