内容分析方法
1
目录
内容分析方法的定义
研究方法的演变和发展
内容分析方法的研究对象
内容分析方法的方法论基础
经验研究方法的分类——定量和定性内容分析
经验研究方法的基本流程
需要分析的三个单位：信息、信息的发送者和接受者
信息的构成和研究分析的基本单位
对信息编码的三种主要方法
利用已有编码和开发自主编码
开放式编码的四个原则
信息的发送者和接受者
内容的抽样
定量和定性内容分析
分级技术：非结构分析和结构分析
配对比较、等级化和直接划分
信度和效度检验
结论
内容分析方法的定义
顾名思义, 内容分析法(content analysis) 是一种对研究对象的内容进行分析, 透过现象看本质的科学方法。
内容分析法是对文献内容进行系统、客观地定量分析的科学研究方法，以分析特定词语或概念在多文本或多个文本集中出现与否、涵义及其相互联系为主；其目的是弄清楚或测验文献本质性的事实或趋势，揭示文献特有的隐性情报内容，对事物发展作情报预测；内容分析方法实际是一种半定量研究方法，其做法是把媒介上的文字、非量化的有价值的信息转化为定量的数据，建立有意义的类目分解交流内容，并以此来分析信息的某些特征。因此，内容分析法是基于定量研究的定性研究方法
研究方法的演变和发展
理论发展
内容分析法最早见于18 世纪的《锡安歌集》之争, 19 世纪美国报业崛起, 内容分析法广泛用于传播领域。第二次世界大战期间的美国学者H1D1 拉斯维尔等人组织了一项名为“战时通讯研究”的工作, 以德国公开出版的报纸为分析对象,获取许多军政机密的情报, 这项工作不仅使内容分析法显示出明显的实际效果, 而且在方法上取得一套模式。20 世纪50 年代美国学者贝雷尔森发表《传播研究的内容分析》一书, 确立了内容分析法的地位。真正使内容分析法系统化的是J ·奈斯比特, 他主持出版的“趋势报告”就是运用内容分析法。享誉全球的《大趋势———改变我们生活的十个新方向》一书就是以这些报告为基础写成的。
研究方法的演变和发展
实践
内容分析法作为一种研究社会现实的科学方法, 经过了不断的理论探讨和实践应用才逐步趋于成熟与完善。主要分为以下几种方法类型: ①解读式内容分析法(hermeneutic content analysis) , 解读式内容分析法是一种通过精读、理解并阐述文本内容来传达作者的意图的方法; ②实验室内容分析法(empirical content analysis) , 实验室内容分析法主要是定量内容分析和定性内容分析相结合的方法; ③计算机辅助内容分析法( computer - aided contentanalysis)。
内容分析方法的研究对象
内容分析法的研究对象为信息的内容特征，他用比较规范的方法读取信息内容，并将大量的信息特征有序地、量化地表达出来。内容分析法的对象包括显性信息和隐性信息
随着计算机技术的发展，内容分析法的研究对象从村文本信息扩大到图像、声频、视频、多媒体等信息，研究的重点从显性只是的同解放南溪转移到了对隐性知识的萃取提炼。
内容分析方法的方法论基础
推理方法
趋势推理（trend inferences）是一种纵向推理方法，是一种分析表示某一特征的信息的数量、重要性、强度等指标在不同时许丽的变化和差异的方法，比如，通过对近年来国内报刊中对创业精神的报道及争论的分析，推断这一转变的发展趋势
共变推理（covariation inferences）：一种通过对表示两个以上时间的信息同时出现的情况进行推断分析，得出关于这几个事件之间的相关性结论的方法。如对企业创业家职业生涯与国家教育发展的相关性分析
因果推理（causal inferences）一种从表示特定事件的文字符号、词语、语句登特则很难给内容的变化来推断事务发展变化趋势的方法。如功过社论推断政治环境。
内容分析方法的方法论基础
比较方法
趋势比较（trend comparison）：即历时比较。他强调同一事物在不同时期的变化内容的比较，从事务相关信息的时序变化中把握事物的发展规律。趋势比较要确保一定时间段，如5，10年。采样的长短往往根据研究需求而定，一般是以年为单位，耳且至少要才几年的信息，分析才有比较的意义。
不同内容群的比较（comparson of diffrernt bodies of content），即围绕一个主题，比较来自不同信息源的内容，从而得出结论。比如，海湾危机船后，可以比较各国大报大刊对这一事件的反应，从而推至各国的立场
内容内比较（intra-content comparison）这种方法是对同一文献中不同主题进行比较，以揭示他们之间的相关关系，反映出同一信息源对不同事件的反应。如西方学者通过分析《纽约时报》对白人和黑人的用词，得出意识形态方面的结论
有标准的内容比较（comparison of content with astandard）是一种以一定标准作为尺度，对同类的信息进行相应的内容比较的方法。所制定的标准可以是抽象的，也可以视具体的，如，在选定国外报刊时，对其公正性、客观性进行评价时就可以制定比较抽象的分析标准
研究方法的分类
两种不同的类型
（1）定量内容分析
（2）定性内容分析
定量内容分析
特征：
频数分析为主 ，进行描述性假设检验
经常使用于结构化的访谈和问卷过程中
定义编码结构
编码体系的不足
如何进行定量内容分析
原始访谈或问卷的分析
固定的编码结构开发与应用
定量化的变量与定量分析
定性内容分析
目标：
“为了系统化地识别交流中的具体特征和有目的地将原始材料转化为科学数据”
——Mostyn, 1985
定性内容分析
数据分类
建立研究主题——研究的具体问题
概念化
更高层次的概念范畴
潜在内容
定性内容分析
持续的发现过程
不断地比较
通过访谈可能会发现研究问题
为什么进行定性内容分析
缩减不必要的数据
进行解释性的分析——一般性分析与分类
对定量化数据进行解释
新的理论体系或措施建构
对预定假设进行检验
对关键问题进行识别
问卷的开发
对已知问题的回答
定性内容分析的选样
根据研究的问题进行选样
选样分为：理论性不高的问题
高度理论化的问题
如何进行定性化内容分析
开放式编码
主题和分类
关系的识别
高层次概念或次级主题
不断的发现——就是不断的比较
编码文本的内容 (类似行为编码)
采用各种推理技术识别信息（语言、文字或影像）的特殊特征
信息被精炼或分类，并在一个编码体系的基础上进行比较
2
研究对象：任何信息材料
素材来源于参与观察的纪录、信件、小说、沟通纪录或影响（如电视节目、访谈等）
3
内容分析的步骤:
充分描述研究的现象 (例如媒介早期的描述portrayal ）
确定研究的主题和研究目的
选择数据合适的媒介（media）
推导出编码类别（ coding categories）
选择类别：特征状态、心理特质或上下文的关系等
计算（ count）已存在或不存在的编码类别
选择其中一种编码类别进行细分（强制性选择）
决定抽样策略—我们不可能计算全部！
培训编码员/评估员 (信度是最为重要的）
分析数据 (%’s, compare means and variances?)
4
例子：美国学者拉斯维尔等人 (1949)
二战期间使用的符号分析技术
报纸的内容作为研究对象，如犹太人、斯大林、民主主义和俄罗斯
一个有感情色彩的概念维度被增加进来——纵容、支持和中立等
有时根据强调程度被区分为不同的级别，如仁慈、道德、软弱和不道德等
当今这些工作都可使用计算机完成
抽样策略
仅仅选择报纸
根据字母排列，以10为等距抽样
地理位置、政治和经济导向和民族作为控制变量
5
通常我们必须决定——
哪些单位需要分析（单词还是图片）
在什么层面上进行分析（类别还是数量？）
在那些材料上进行编码（每张报纸的第10页还是其他所有语句）
6
信息的基本构成
信息,
信息的发送者,
信息的接受者。
7
信息的构成
清楚的主题,
不同论题的强调
话题的时间和空间量
等等.
8
对文本进行计量:
计数（counting） vs. 分级（rating）
计数—判断性不强，发现性也不强
little judgment, less discovery
分级—判断性强，发现性也强
rating--much judgment, more discovery
通常两个办法综合使用
9
count angry words in text or rate anger in sentences or paragraphs
“count” angry words, what is an angry word???
end up using some judgment unless it’s totally spelled out
and the more it’s spelled out, the more you lose discovery
10
. Anger
count these words:
fed up IIII II
irritated III
disgusted II
etc. IIII III
or rate these sentences:
I find the idea to be distasteful.
very warm very angry
0 1 2 3 4 5
11
examples of elements to count:
items,
words
sentences
paragraphs,
characters,
semantics
concepts,
themes,
13
Items
represents the whole unit of the sender's message
may be an entire book, a letter, speech, diary, newspaper, or even an in-depth interview.
14
Words.
smallest element
least judgment
generally results in frequency distributions
15
Sentences
definitely more judgment than words but less than paragraphs, etc.
16
Paragraphs
difficulties result in attempting to code and classify
poor consistency among writer on how to write a paragraph
17
Semantics
meanings
of overall sentence, paragraph, etc.
requires a lot of judgment
18
Characters (persons)
count the number of times a specific person is mentioned
19
Concepts
involve words grouped together into conceptual clusters (ideas)
. a conceptual cluster may form around the idea of deviance.
Words such as crime, delinquency, money laundering, and fraud might cluster around the conceptual idea of deviance
leads toward more latent than manifest content, more rating, judgment, etc. although word clusters could be simply counted
20
Themes
broader than a concept (almost like a mood)
can be made up of many concepts
must further specify the unit --theme of each sentence, each paragraph, the whole book????
21
Combinations of Elements
. Berg's (1983) subjective definitions for Jewish affiliational categories
22
used combination of word, sentence, and paragraph elements
lifted definition components from interview transcripts cutting across elements to formulate interviewee’s definition
each definition annotated with the transcript number
23
example--Interview #60:
ORTHODOX
Well, I guess, Orthodox keep kosher in [the] home and away from home. Observe the Sabbath, and, you know..., actually if somebody did [those] and considered themselves an Orthodox Jew, to me that would be enough. I would say that they were Orthodox.
CONSERVATIVE
Conservative, I guess, is the fellow who doesn't want to say he's Reform, because it's objectionable to him. But he's a long way from being Orthodox.
REFORM
Reform is just somebody that, say they are Jewish because they don't want to lose their identity. But actually want to be considered a Reform, 'cause I say I'm Jewish, but I wouldn't want to be associated as a Jew if I didn't actually observe any of the laws.
NONPRACTICING
Well, a Nonpracticing is the guy who would have no temple affiliation, no affiliation with being Jewish at all, except that he considers himself a Jew. I guess he practices in no way, except to himself.
24
Three major approaches to categorizing in a coding system:
common classes,
special classes, and
theoretical classes
25
1. Common classes.
used by virtually anyone in society
(for example, age, gender, mother, father, teacher, boss, lover, etc.)
essential in assessing whether certain demographic characteristics are related to patterns that arise from other coding
26
2. Special Classes.
colloquial categories
includes jargon of various professions, . petty larceny vs. that other category
27
3. Theoretical Classes.
those that emerge in the course of analyzing the data
category labels generally borrowed from special classes (. psycho jargon)
their substance is grounded in the data
not immediately knowable until observers spend considerable time with the content
28
"in vivo codes" vs."sociological constructs"
in vivo codes are literal
actual words
sociological constructs are formulated by the analyst
. "professional attitude," "family oriented," "obsessive workaholic," "educationally minded," might represent examples of sociological constructs
may add breadth and depth
29
In General,
you can apply an established code but
you lose some of the individuality in the data set
or you can develop your own code but
people may accuse you of post hoc or circular reasoning in this case
45
how to identify categories in the data and apply them--at the same time without being “too circular” in your reasoning
OPEN CODING
here’s a quote--don’t copy it
30
“Inexperienced researchers, although they may intellectually understand the process described so far, usually become lost at about this point in the actual process of coding. Some of the major obstacles which cause anguish include the so-called true or intended meaning of the sentence, and a desire to know the real motivation behind a clearly identifiable lie uttered by a subject. If the researchers can get beyond such concerns, the coding can continue. For the most part, such concerns are actually irrelevant to the coding process, particularly with regard to open coding, the central purpose of which is to open inquiry widely. Although interpretations, questions, and even possible answers may seem to emerge as researchers code, it is important to hold these as tentative at best. Contradictions to such early conclusions may emerge during the coding of the very next document. The more thorough analysis of the various concepts and categories will best be accomplished after all of the material has been coded” (Berg, 1990).
31
“Inexperienced researchers, although they may intellectually understand the process described so far, usually become lost at about this point in the actual process of coding. Some of the major obstacles which cause anguish include the so-called true or intended meaning of the sentence, and a desire to know the real motivation behind a clearly identifiable lie uttered by a subject. If the researchers can get beyond such concerns, the coding can continue. For the most part, such concerns are actually irrelevant to the open coding process, the central purpose of which is to open inquiry widely. Although interpretations, questions, and even possible answers may seem to emerge as researchers code, it is important to hold these as tentative at best. Contradictions to such early conclusions may emerge during the coding of the very next document. The more thorough analysis of the various concepts and categories will best be accomplished after all of the material has been coded” (Berg, 1990).
32
“Inexperienced researchers, although they may intellectually understand the process described so far, usually become lost at about this point in the actual process of coding. Some of the major obstacles which cause anguish include the so-called true or intended meaning of the sentence, and a desire to know the real motivation behind a clearly identifiable lie uttered by a subject. If the researchers can get beyond such concerns, the coding can continue. For the most part, such concerns are actually irrelevant to the coding process, particularly with regard to open coding, the central purpose of which is to open inquiry widely. Although interpretations, questions, and even possible answers may seem to emerge as researchers code, it is important to hold these as tentative at best. Contradictions to such early conclusions may emerge during the coding of the very next document. The more thorough analysis of the various concepts and categories will best be accomplished after all of the material has been coded” (Berg, 1990).
33
“Inexperienced researchers, although they may intellectually understand the process described so far, usually become lost at about this point in the actual process of coding. Some of the major obstacles which cause anguish include the so-called true or intended meaning of the sentence, and a desire to know the real motivation behind a clearly identifiable lie uttered by a subject. If the researchers can get beyond such concerns, the coding can continue. For the most part, such concerns are actually irrelevant to the open coding process, particularly with regard to open coding, the central purpose of which is to open inquiry widely. Although interpretations, questions, and even possible answers may seem to emerge as researchers code, it is important to hold these as tentative at best. Contradictions to such early conclusions may emerge during the coding of the very next document. The more thorough analysis of the various concepts and categories will best be accomplished after all of the material has been coded” (Berg, 1990).
34
IF YOU DEVELOP YOUR OWN CODE
YOU LET THE DATA SPEAK FOR THEMSELVES
46
GROUNDED THEORY IS
an interplay of experience, induction, and deduction
1. initial systematic discovery of the theory from the data
2. application of the theory to the data
47
Remember this--
FACT
THEORY
0 0 0 0 0
revision
0
deduction
induction
48
NOW APPLY IT TO CONTENT ANALYSIS
FACT
THEORY
READ TEXT
revision
0
deduction
induction
READ
AGAIN
(expect something)
49
Researcher examines newspaper orientations toward changes in seatbelt law!
First, read each article and ask which ones favor and which against
Was the decision to label an article pro or con based on the use of certain terms, on presentation of specific study findings, or because of statements offered by particular characters (who?)?
Were the article's positions more clearly indicated by their manifest content or by some undertone?
96
answers to these questions lead to inductive categories in which to slot various units of content
and deductive application of the categories leads to more inductive categories
97
“If investigators have begun with specific empirical observations, they should attempt to develop explanations grounded in the data (grounded theory) and apply these theories to other empirical observations (deductive reasoning).”
“If they are attempting to test theory derived from previous research and previous inductive reasoning, their theoretical orientation should suggest empirical indicators of concepts (deductive reasoning).”
Quotes
(don’t copy)
98
specification of the
content characteristics
application of rules
for counting those
characteristics
revision
0
deduction
induction
o o o o
o
Remember the hermaneutic circle or the hypothetico deductive approach---
99
One way to avoid the accusation that you’re engaging in post hoc or circular reasoning is to:
divide the data set in half
develop the code on one half
apply it to the other half
50
Another way
totally replicate the study
“you think I made it up or imagined it, watch I’ll do it again!”
51
SO GROUNDED THEORY IS AN:
interaction of two processes:
specification of content characteristics
application of explicit rules for counting those characteristics
52
specification of the
content characteristics
application of rules
for counting those
characteristics
revision
0
deduction
induction
o o o o
o
As you apply the coding rules, you often find more characteristics to code---OR LESS (SEE NEXT EXAMPLE)
DISCOVER MORE
CHARACTERISTICS
53
identified several themes on intake sheets
set up a tally sheet,
start to categorize criminal offenses declared by arresting officers
discovered two distinct classes of crime
1. shoplifting , petty theft, and retail theft all referred to stealing of some type of store merchandise, usually not exceeding $ in value
2. the similar term petty larceny was used to describe the taking of cash whether from a retail establishment, a domicile, or an auto
The intake sheets at first suggested 4 categories (shoplifting, petty theft, retail theft, petty larceny)
Recent study evaluates effectiveness of Florida delinquency program!
54
Strauss (1987) says-- "believe everything and believe nothing" while undertaking open coding
he lists the following 4 guidelines:
35
1. Ask the data a specific and consistent set of questions.
most general
What study are these data pertinent to?
what was the original objective of the research study?
original purpose of a study may not be accomplished, and an alternative or unanticipated goal identified in the data.
36
For example,
Pearson (1987) --evaluation of a New Jersey intensive probation supervision program-- original aim = demonstrate cost effectiveness
objective indicators failed but indirect indicators suggested success
reports from relatives of probationers about changes in attitudes
spouse reported that her husband now sent child-support payments
parent reported that child now showed personal responsibility by doing household chores
37
Thus, Pearson (1987) points to an unanticipated benefit from the program.
keep the original study aim in mind and remain open to multiple or unanticipated results that emerge
38
Strauss (1987) also says--
2. Analyze the data minutely.
in the beginning, more is better
like a funnel
the wide end = many categories, incidents, interactions, etc.
coded minutely during open coding
let patterns emerge or reveal themselves
39
Later,
more systematic coding can be accomplished, building from the numerous elements that emerge
The question is when to stop open coding and move on to the speedier, more mechanical coding phase?
Typically, as researchers minutely code, they eventually saturate the document with repetitious codes.
as this allows researchers to move more rapidly through the documents, it is usually time to move on.
40
3. Strauss also says frequently interrupt the coding to write a theoretical note.
called ‘grounded theory’
a comment in the document triggers ideas--jot down a note
otherwise it’ll be forgotten
keep a record of where in each document similar content triggered the idea
41
For example,
study on adolescents' involvements with alcohol, crime, and drugs revealed youths speaking about drugs and criminal activities as though they were partitioned categories (Carpenter et al, 1988). Notes scribbled during coding later led to theories regarding youth’s drug-crime culture.
42
4. Finally Strauss says that in open coding never assume the analytic relevance of any traditional variable such as age, sex, social class, and so on, until the data shows it to be relevant.
even these more mundane variables must "earn their way into grounded theory"
What are the study data pertinent to?
What are you trying to find out, trying to say??
43
If researchers are interested in gender differences, they begin by assuming that gender might be analytically relevant, but if the data fail to support this assumption, the researchers must accept this result.
44
资料整理与分析
定性资料分析运用的是归纳法，即通过整理分析资料得出假说或理论；
定量资料分析运用的是演绎法，即事先定好一个理论或假说，然后收集资料以验证这一理论或假说是否成立。
资料整理与分析
首先将原始资料系统化、条理化，然后将资料编码、分类、集中、比较和浓缩，最后对资料进行解释、推广、形成假说、理论或基础理论。并再次在原记录中对这些理论或假说进行验证、修改，这一过程有时需要反复多次才能完成。这一整理分析过程因研究目的和研究问题不同而不同，没有一套固定的程序。
有时这种分析过程开始于资料的收集阶段，这与定量资料的分析（分析是在资料收集结束之后开始的）明显不同。
对偏离常规的观念进行分析是重要的，即必须注意少数派的观点及那些与研究者总的理论不一致的例证。
资料整理与分析
分析专题小组讨论会资料的特殊之处：
将同一主题的小组讨论会资料集中起来并比较，然后研究不同的观点如何来自不同的样本人口。重要的是区分哪些是个人观点、哪些是小组多数人的观点。
需要标明群组动力学的影响并分析专题讨论会在哪些方面充分利用了研究参与者之间的互动。
基本分析程序
逐字逐句、认真细致地阅读原始资料，对具体内容进行分类、编码，对不同的文件赋予相应的属性，并在备忘录上记录自己的一些初步想法等。然后从所有自杀访谈中抽出某一类别的内容再认真阅读、思考，并与其它类别的内容进行比较、联系，形成相应的理论，再回到访谈资料中验证理论是否正确，若有出入则对理论进行修正。如此反复多次，得出结论。
资料分析
运用“QSR Nvivo”定性分析软件对内容进行分析：前提是将录音资料全部以文字形式逐字逐句输入电脑
编码（Code）：即用恰当的概述性文字对文字内容的某一部分做出标记。
自由编码（FREE NODES）
树状编码（TREE NODES）
案例编码（CASE NODES）
自由编码（FREE NODES)
在进行文字分析时，通过反复阅读原始资料，最终确定使用的编码如下：
当地的自杀原因
当时如何避免自杀行为的出现
当时是否采取措施避免被人发现
对自杀者的看法
家人自杀是否需要保密
家庭环境如何？
留遗嘱
哪些自杀值得或可被理解
如何预防当地人的自杀
是否认为死比活着好
是否听说或见过自杀案例
与朋友、邻居的关系如何？
与亲戚的关系如何？
自杀当时如何考虑被救活的可能性
自杀当时如何考虑此方式的致死性
自杀目的
自杀前的工作状况如何？
自杀前的饮酒状况
自杀死后躯体和灵魂会怎样？
树状编码（TREE NODES）
自杀过程（1）
自杀原因（11）
自杀前是否想过用其它解决办法（12）
有无亲朋可以帮助解决（13）
想过自杀能否解决问题（2）
自杀当时（21）
现在（22）
现在问题有无好转（23）
自杀意念（3）
出现时间（31）
如何决定用此方式自杀的（32）
当时是否有人知道你有自杀意念（33）
自杀的影响（4）
对家人（41）
他人的看法有何改变（42）
对定性研究方法的评价
定性研究方法并不容易，其资料收集、整理和分析既繁琐又复杂。这种资料收集方法并不比其它方法收集的资料更加或更少真实，但对于研究某些特殊类型的问题却是最恰当的方法。
质量评估
争议的焦点：能否用定量研究的评价标准如信度、效度和可推广性来评估定性研究的质量
比较一致的倾向：用效度来评估定性研究的质量，即能否反映人们真实的经历、态度、行为与动机等。验证效度常用三角检验法，即用三种或更多种方法（如一个大型的量表调查、专题小组讨论和一段时期的观察）对某一专题进行研究并对结果进行比较、验证或补充。但也可评估其信度，即通过比较不同人员独立分析同一资料的结果是否一致来验证。
定性分析软件的
作用
与手工分析相比，分析软件仅能免去记笔记、剪贴的麻烦，而不能代替人脑进行分析，仍需研究者逐字逐行地深入阅读分析。
适合的研究领域
观察法：适合于研究社会角色和正式组织
访谈：适合研究个体经历
专题小组讨论：更适合研究态度和经历，以及在特定的文化背景下知识（特别是想法）是如何产生的及如何运用的。如当研究反复发现健康知识和健康行为之间存在鸿沟时，只有定性方法如专题小组讨论，可以填平这些鸿沟并解释为什么会出现鸿沟。
All of our discussion of
COMMUNICATION COMPONENTS
message
sender,
audience
has centered on “message”--let’s turn to sender
55
MESSAGE VS. SENDER
What is said?
versus
How is it said?
57
sender
linkages between message and attributes of the sender are often slight but
can gain some impression from numerous examples,
. recordings, transcripts including literal representations of pauses, mispronounced words, grammatical errors, slang, etc.
56
now the last --COMMUNICATION COMPONENT
message,
sender,
audience
58
audience
analyze content from the audience as well as from the sender of the message
. Pornography and Television Violence Commissions
write a description or essay about the program you just watched or
interview and content analyze the interview transcription
Also try to get a feel for who is consuming the message
when I see ads for feminine napkins on a program I assume the network has calculated that a lot of women will be watching the program--similarly its aimed at men if shaving products are emphasized
59
Sampling Content
In general, content analysis samples:
1. sources (newspapers? which ones? or magazines? which ones?
2. dates of sources (time periods)
3. units or chunks of content within the source (headlines? editorials? cartoons?)--or you can simply divide the newspaper into inches, lines, words, quadrants, and sample from within these units
60
criteria of selection of sources and components:
should reflect all relevant aspects of the messages and retain, as much as possible, the exact wording used
quote: “not merely the arbitrary or superficial application of irrelevant categories.”
a priori operational definition of content to be included and excluded
try to eliminate analysis in which only content supporting investigator's hypotheses is examined
categories also emerge in the course of developing these criteria
61
SAMPLING STRATEGIES (applicable to all forms of research)
1. simple random sampling,
2. systematic sampling,
3. purposive sampling
62
1. Simple Random Sampling
rely on chance to generate a representative sample
draws subjects from an identified population in such a manner that every unit has precisely the same chance (probability) of being included in the sample
63
Sampling and Generalizing
64
Sampling
The Population
The Sample
65
Sampling
The Population
The Sample
This
represents
this
66
Generalizing
The Population
The Sample
This is used to
make statements
about
this
67
If you pick only the prettiest flowers from your garden,
68
If you pick only the prettiest flowers from your garden,
is that a random sample?
SAMPLE
GARDEN
69
If you pick only the prettiest flowers from your garden,
is that a random sample?
Would the flower you picked
represent your garden?
70
2. Systematic Sampling
get a printed list of population members
sample every Nth name
sampling interval = divide number of people desired into the full population size
. sample of 80 from a population of 2560 means you sample every 32nd person (2560/80 = 32)
be sure to start at a random point in the list
71
One type of systematic sampling = Stratified Sampling
ensure that a certain segment of the population is correctly represented
population is divided into subgroups (strata) and the % of the sample taken from each stratum equals its representation in the population
. if SIU has 55% men and 45% women, that’s the proportion we put into our sample
72
3. Purposive Sampling
researchers use their special knowledge or expertise to select subjects who represent the population
. use the Eysenck Personality Inventory to select neurotic subjects for your sample
must be careful generalizing
73
Example of sampling in content analysis: Graber (1971)
She developed a complex sampling scheme for examining newspapers
country divided into regions
cities in each region divided into population size groupings: over a million, a million to half a million, and fewer than half a million
3/4 of the sample drawn from the most populous states in each region to reflect “voting power”
half were selected from states in which Democratic party was dominant and half from those where Republican party was dominant
representative sampling of monopoly vs. competitive newspapers
then all campaign stories were coded
74
Quantitative vs. Qualitative Analysis
hot debate over whether analysis should be quantitative or qualitative
Quantitative = tally sheets, specific frequencies of relevant categories
Qualitative = ratings, examine ideological mind sets, themes, topics, symbols
while “grounding” such examinations in the data
75
REMEMBER LEVELS AND UNITS OF ANALYSIS (this sometimes affects the extent to which you can be quantitative)
levels of prose:
words, phrases, sentences, paragraphs, sections, chapters, books, writers, ideological stance, subject topic, etc.
television programs:
segments between commercials, entire program, entire prime time periods, etc.
76
one side says analysis should be "objective, systematic, and quantitative."
but quantitative analysis emphasizes "the procedure of analysis," rather than the "character of the data available"
77
also “quantitative content analysis results in a somewhat arbitrary limitation in the field by excluding all accounts of communications not in the form of numbers, or those which may lose meaning if reduced to a numeric form (definitions, symbols, detailed explanations, and so forth)”
78
Other proponents, . Smith (1975), suggest that some blend of both quantitative and qualitative analysis should be used.
79
Smith (1975: p. 218) explains that he has taken this position "because qualitative analysis deals with forms and antecedent-consequent patterns of form, while quantitative analysis deals with duration and frequency of form."
80
Another hot debate--
Manifest versus Latent Content Analysis
manifest content (those elements that are physically present and countable)
latent content (interpretive reading of the symbolism underlying the physically presented data)
. examine an entire speech to rate how "radical" it was, or a novel could be considered in terms of how "violent" it was
81
manifest content = "surface structure,”
latent content = "deep structure" (meaning)
82
Holsti (1969) has tried to resolve this debate:
“It is true that only the manifest attributes of text may be coded, but this limitation is already implied by the requirement of objectivity. Inferences about latent meanings of messages are therefore permitted but . . . they require corroboration by independent evidence”...
83
also “should offer detailed excerpts from relevant statements that serve to document the researchers' interpretations” ...
“A safe rule of thumb to follow is the inclusion of at least three independent examples for each interpretation”
84
Blending Manifest and Latent Content Analysis
Take note of apparent presence of a concept and
report the frequency at which an indicator appears to suggest or reflect magnitude but
be cautious about generalizing to actual magnitude
., If indicators of "positive attitudes toward shoplifting," appear fifty times in one, and twenty-five times in another, do not claim that the first is "twice as likely to shoplift."
stay close to the “facts”
85
Content Analysis is continued with some examples and discussion of rating techniques in the next section of the course
: )
信度
前人的工作
多个编码人员
编码人员之间的信度（定量内容分析）
信度＝N×平均相互同意度／（1＋（N－1）×平均相互同意度）
效度
三角测量
预期效度
参与者验证
研究的透明度
Example of Content Analysis
Davis (1984)--sexism in children’s books
Generally, wanted to discover if children’s books specifically written to be nonsexist were nevertheless sexist with regard to both male and female roles
96 children’s books from three categories
written to be nonsexist
award winning books
best sellers
86
Davis developed 15 categories
including
independent
self-initiating and self-sustaining
active
physical activity, work, play
passively active
fine motor physical activity such as reading, talking, and thinking
emotional
affective displays of feelings
87
Four graduate students worked as raters
to establish interrater reliability
2 worked only on illustrations
2 worked only on text
88
Results
note that you can compare male and female within a type of book (. bestsellers) or across types of book (. bestseller vs. nonsexist)
89
Results-- active
90
More results--aggressive
91
more results--emotional
92
more results--submissive
93
more results--nurturant
94
maybe these books are written for girls--maybe they read more--as a matter of fact look at Passively Active (includes reading)
95
RATING TECHNIQUES (TRICKS OF THE TRADE)
1. Q-SORT
2. PAIR COMPARISON
3. RANKIING
4. DIRECT SCALING
1
1. Q-SORT TECHNIQUES
Statements are put on note cards to form a “deck” (like playing cards)
Unstructured Q-Sort
raters are asked to sort the cards anyway they want
raters have to invent categories
Structured Q-Sort
raters are told what categories the cards are to be sorted into
the categories come from a theory
the arrangement of piles helps illustrates the meaning of the categories (see next slide)
2
low
high
sort these
SIMPLE
MAGNITUDE
EXAMPLES
3
CAN GIVE RATERS SETS AND THEN SUBSETS TO SORT
EXAMPLE FROM PREVIOUS SLIDE
SORT THESE
INTO TWO PILES
SORT THESE
INTO TWO PILES
SORT THESE
INTO TWO PILES
4
Analysis unstructured--
5
what categories do they use
and if they use some of the same categories, how much do they agree
% AGREEMENT (remember A/A+D?)
differences in categories used by different types of raters
groupings of raters based on the categories they use
similarity of ratings to an ideal set of categories generated by a theory
6
Analysis structured--
7
structured--If sorting by magnitude
mean rating and the variance of the ratings for each stimulus
differences in rating between different types of raters (. men vs. women might rate them differently)
similarity of ratings to an ideal generated by a theory
groupings of raters based on their ratings (grouping) of the cards
8
structured--if sorting hierarchically
how much do they agree
% AGREEMENT
plus all of the above analyses
9
Other rating methods:
10
2. method of pair comparison
every stimulus is paired with every other stimulus
results in N(N-1)/2 pairs ( 10 stimuli give you 45 pairs)
1 2 3 4 5 6 7 8 9 10
1
2
3
4
5
6
7
8
9
10
45 pairs but if you
balance the order (which
one they see first) then
you have 90 pairs
11
You’ll be asked to pick the most beautiful, or the most angry, or etc.
Two at a time over and over
12
Example with statements-
13
Which statement is most offensive to women?
Have you ever seen a women mechanic?
Women should never have been given the right to vote.
Women’s athletics should be given less money than men’s athletics
Have you ever seen a women mechanic?
Women’s athletics should be given less money than men’s athletics.
I would never go to a woman surgeon.
14
Example with pictures-
15
Which painting is most sexist?
A
B
16
statistics for pair comparison
it’s like a democracy--how many votes does each stimulus get?
A____ vs. B____
A____ vs. C____
B____ vs. C____
etc etc
17
Lengthy example--try to feel what it is like to do the rating.
18
Which person is the most religious?
A
B
19
Which person is the most religious?
A
B
20
Which person is the most religious?
A
B
21
Which person is the most religious?
A
B
22
Which person is the most religious?
A
B
23
Which person is the most religious?
A
B
24
Which person is the most religious?
A
B
25
now we start to shuffle the pairings
26
27
28
29
30
31
YOU WOULD CONTINUE IN THIS WAY UNTIL EVERY ONE OF THE 16 PICTURES HAD BEEN PAIRED WITH EVERY OTHER ONE
ANOTHER QUESTION --WHAT ABOUT THE ORDER OF PRESENTATION??
SHOULD WE PRESENT EVERY PAIR TWICE, ONCE WITH A PICTURE ON THE LEFT AND ONCE ON THE RIGHT???
32
Other methods--
33
3. method of ranking
like this
but different in that there are no categories to put them in---just rank order them (several times)
then average the ranks
34
4. method of direct scaling
apply a rating scale to each stimulus one at a time
example: this statement “I was hoping that you might help me with this problem” is
not asssertive
at all
very assertive
1 2 3 4 5
35
Method of Direct Scaling
RATE HOW RELIGIOUS THIS MAN IS:
1 2 3 4 5 6 7
36
RATE HOW RELIGIOUS
THIS MAN IS:
1 2 3 4 5 6 7
NOT
RELIGIOUS
AT ALL
VERY
RELIGIOUS
37
38
39
40
AND FINALLY
THE MOST IMPORTANT QUESTION OF ALL:
WOULD YOU BUY A USED CAR FROM THIS MAN???
41
Limitations
Bias
Group level interpretation - generalised themes may not apply to individuals
Descriptive - not theory generation
End of Content Analysis
1
2
3
4
5
6
7
8
9
10
11
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
45
30
31
32
33
34
46
47
48
49
96
97
98
99
50
51
52
53
54
35
36
37
38
39
40
41
42
43
44
55
57
56
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
