《如何編制優質的問卷》书评

笔记

问卷调查提问：other-administrated，self-administratedstructured(early exploration stage), unstructured(controversy, the later stage)策略：amount of structure, degree of directnessadaptive tailored questioning->Bayesian information seeking approach内容告知：full disclosure, indirect -> red herrings（事后报告）informed consentQUESTIONNAIRE IS A LEARNING DEVICE“uninformed responses. 并不了解主题，但无论如何都能提供答案”QUESTIONNAIRE ANSWERS ARE DATA1.矛盾、反应、可取定理(Zallar&Feldman)The source of an answer: long-term memory? Maybe not.了解答案可能是如何形成的2.The mode of question administration.3.研究对象的认知与沟通技能CH2 问卷编制与研究相关（怎么分析，可与其资料，如何使用？）优先顺序 研究对象理解，有能力回答，愿意回答 (personal, too much effort or time)literal meaning->absolute understanding.relative meaning->the gray-mite meaning.“Garbage in, Garbage out!”CH3 开放，封闭content analysis. space-tole. easy-to-answer 开关，建立关系 verbosity effect封闭 小于9个选项 (If you ask a question, you will get an answer, regardless of how meaningful this question is.) Bullet effect.不要问价格。dummy ringer sleeper control.互相独立，包罗万象 (mutually exclusive, collectively exhaustive)CH4the precise wording. 有时empirical testing无法设计完美的问题，but avoid bad or poor question:(1)formulate a question as you would in normal conversation. common sense, the mode of questionnaire administration context. (常识)(2)QAP: question-answer paradigm(3)readability formula(经验)判断基准：（有例外）BRUSO: brief, relevant, unambiguous, specific, objectiveBrief. KISS every question, every word in a question: direct cost - survey cost, opportunity cost - design cost, indirect costs - may refuse to answerRelevant. (perceive) pretest. relevancy vs. repetitiveUnambiguous. multiple meanings.->unwanted variability in question answers.(slang, jargon...专业术语 罕用 避免) 多重意义：”cold weather”=?C “middle age”=?years oldcould?should?might?country?energy? 开放式更易模糊不用vague quantifiers：near, very, quite...<量化定调> <self-perceptions=>模糊的也有意义> 用字典和百科全书确保词义Specific. 一次只问一个问题。 last month而非an average month. 避免绝对: ever, always, never. 避免一般性：generality.cheater question “题目隐瞒某些资料而取得优势？”->找出随意回答问题的人Objective. 无暗示：no-win question. leading question.事先提出一种共同想法的假设隐约强调一个议题的某一面向带有社会期待的条件制约（人们避免冒犯）social desirability bias提供研究对象充分的资料减少一个问题的威胁性质，先提供研究对象一项有关民众的资料。CH5 等级量尺与量表规则必须isomorphic & non-degenerating. (one-to-one correspondence)单向度量尺 uni-dimensioned scales = comparative scales+zonaclic scales选项个数 7+－2注：limited scaling capabilities. 无法细分 南加州人与北印第安人对冰的描述言语刺激(verbal stimuli): comparison(更多－更少), endorsement(确定有－确定非有), frequency(总是－从不), influence(主要－次要), intensity(非常－非常不)unipolar - bipolar(检查是不是反义字？)定点越强，变异越少！->pretest 选项平衡类别选项的标记 按照惯例，参考之前的。repertory grid technique. focus group interviews.“不知道或没意见”项 a comparative rating scale.the semantic differential scale，the Likert scale，the stapel scale (less used)评价uni-dimensional(interned consistency, longitudinal stability)reliablevalid(content, face, surface validity)generalizableCH6提问特定的问题人口统计学放最后 status effect. 收入：假造最低项，最高项数造低些回溯性报告bogus pipelineerrors of omissionerrors of commissionerrors of telescopingCH7 结构&评鉴preface/cover letter：强调问卷的重要性强调研究对象的重要性赞助单位的重要性personalization->salutation间接、提供动机、答案没有对错、不问用户，不给他们说“不”的机会an inducement question(rapport questions)screening or filter questions.substantive question section. 逻辑顺序，由简入难。funnel questions, tri-bi/forkingsgeneral classification questionsthe potential effect of this context on question answers: pure context effect, part-whole question effectthe chameleon effectPretesting a questionnaire 专家群 方便性样本