Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

Content analysis > Chapter 7 > Flashcards

Chapter 7 Flashcards

1
Q

Categories, rules reliability applied > results increase valid
Valid> fact or evidence?
Valid> speakers logic is persuasive

A

-

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Social science

A
  • Breaks reality into distinct parts we believe exist and have observable indicators.
  • Social science operates with logic and properly collected observations to connect those concepts to help predict, explain and control reality.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

CA

A

Scholars must address a concept they defined about communication reality exists in reality / is it the appropriate measurement?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Second issue

A

Linking concepts through data collections and analysis methods with producing successful predictions.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

CA: reality of communication in our world.

A

= creation of reliable and valid categories making up the variables we describe and relate to one another in hypothesis or models of communication processes. Operationalise define content categories to terms of these in hypothesis and questions.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

When asking about validity>

A

Operational definition that reduces ambiguity in measurement of communication reality rather than apprehended reality

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Resolving ambiguity

A

Connecting content measurements to previous research.
Critical problem: efforts to achieve reliability in content category measures.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Measurement reliability

A

Necessary but not sufficient condition for measurement validity.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Measure

A

Can be reliable in application but wrong in really measuring: valid must be both reliable in application and valid for what it measures.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Special issue

A

Reliable measures can come at expense of valid measurement.

Get high levels of coder agreement: operational definition may have only tenuous connection with the concept of interest.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Computer CA:

A

Validity of concepts compromised by the focus of keywords absent any context that gives them meaning. Solution: multiple measures of concept - meaningful beyond the study?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Tests of Measurement Validity

A

Four tests - operational terms we use in our hypotheses and questions. Tests of validity apply to measures, constructs and relationships

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Face validity:

A

Most common, persuasive argument that a measure of a concept makes sense on its face. Obvious to all and no additional explanation.
Good when agreement is high. Enhance face with precious measurements.

Can be chancy - concept can have latent meanings - same concept in different ways.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Concurrent validity:

A

Two different methods and same conclusion. Face validity strengthened: correlate the measures used in one study with a similar in another study. two methods provides mutual or concurrent validation

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Predictive validity:

A

Correlates a measure with some predicted outcome. If outcome occurs as expected: confidence in the validity of the measure increases. Prediction borne out: confidence in the validity of measures making up operational decisions strengthens.

Validating the predictive power of the content model

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Construct validity:

A

Relation of an abstract concept to the observable measures that indicate the concepts existence and change, construct exists but not directly observable except through one or more measures.

Change in underlying concept will occur change in measure.

Stat tests: relate to only that concept and no other?

  • con. validity does not exist: measures may change because of their relation to some other concepts.

When varying: only concept of interest varies. (confident in that.
= present if the measurement instrument does not relate to other variables when there are no theoretical reasons to expect it.

IF researchers find a relation between a measure and other variables predicted by theory and fail to find another one (not predicted by theory) = evidence for construct validity.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Common constructs across studies for coherence and common focus.

A

-

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Validity in Observational Process

A

How to link them in a way: validly describes social reality. Minimise human biases

Random sampling, randomly assigning to create control

In CA: How to use protocol definitions and tests for chance agreement to minimise the influence.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

HOW CA ACHIEVES this

A

Internal and external validity

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Internal validity:

A

Ability of an experiment to illuminate valid causal relations. (controls to rule out influence).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

External validity:

A

Broader relevance of an experiment’s findings to the causal relations in the world. Natural settings in experiments makes this better.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

CA: cannot possess internal causal validity since it cannot rule out all known and unknown third variables. Causal: time order

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

CA: strengthen ability to make casual inferences with CA paired with survey research to explore relations.
CA can be very strong in external validity and generalisability

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

External and notion of social validity=

A

Social significance of content and relevance and meaning.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

Internal Validity and Design

A

CA: illuminate patterns, regularities, or variables relations.
CA alone cannot establish antecedent causes producing those patterns in the content or explain as causal the subsequent effects that content produces in the social system.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
26
Q

CA: should

A

Address issues of control, time order and correlation of variables included in a causal model

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
27
Q

Control in CA

A

Explain patterns of content: look to information outside the content of interest. Theoretical model, factors that may influence content.

28
Q

Control for other influences by bringing them into the analysis.

A
  1. Even a good model forgets things that are left out.
  2. This model can always be empirically tested to assess how well it works to explain patterns in content.
    Unimportant variables can be dropped.
29
Q

Well-thought model identifying concepts in causal process essential to control variables in CA.

A
30
Q

Time order in Content analysis

A

Incorporate time element in design - empirical.

Causation in CA - specification of time order, control and demonstration of joint variation or correlation..time order is built into design.

If content is influenced by antecedent variables: then the antecedent must come first.

If content influences individual, content must be created by the individual

31
Q

Requirements of control and correlation in causality are established statistically.

A

IV and DV influences must be explicit in any kind of multivariate analysis. Consider direct and indirect causal flows.

32
Q

External validity & meaning in CA

A

Is it really meaningful for the application in the world?
Notion of validity: social dimension that relates to how such knowledge is understood.

Must be meaningful for both sender and receiver.
Results from common language, frame or reference when being communicated.

Placed before the scientific community for their assessment of meaningfulness as valid scientific knowledge. Min: validation of the research through peer review = minimise human bias.

Judges: applies scientific criteria for validating the relevance and design method.

33
Q

Why is peer reviewed?
(Peer review is necessary but not to establish a broader meaning.)

A

Comments and criticism for improving the study - illuminate mistakes. and to inform and assist other researchers to predict phenomena.

(Judgment of scientific community provides a link between internal and external validity
research with flaws cannot be trusted.)

34
Q

Scientific validation

A

Research is necessary before research can have broader meaning or relevance.

Internal validity is a necessary (not always sufficient) condition for external validity.

Tentative until replication and extension

35
Q

Even non-probability

A

Useful if it makes a cumulative contribution
Validation can also happen with use, modification and development of studies definitions or measures.

36
Q

External validity:

A

Maximise its social validity.
Social importance of the content and how it has been collected, the social relevance of the content categories and how measures are taken.

37
Q

Issues with social validity=

A

Nature of the content: increased if content is important. Sheer size of audience exposed: exposure of some critical audience to its influence.
Content is important because of its crucial role or function in society.

Whatever importance: social validity will be affected by how the content has been gathered and analysed for the study.

iIf collected through a census or prob will influence how much generalisability

38
Q

Goal in research:

A

Generating knowledge about people, social institutions and documents.
Prob sampling: generalise to the population from sample.

39
Q

Computer code that directs algorithmic coders classification must be UNAMBIGUOUS AND EXHAUSTIVE.
can be more transparent and replicate that human counterparts.

A
40
Q

Human coding:

A

Ambiguous and incomplete protocols

41
Q

ATA (computer to assign numeric values to content based on set of rules)

A
42
Q

If primary algorithmic coder: it is not CA because: it follows different processes and these processes have unique implications for validity of data. Distinct research method: we should apply a separate term to the algorithmic coders.

A
43
Q

Prefer ATA before CATA -

A

Possible to use computer aids to freeze dynamic content for coding, retrieve big amounts of data and sort and organise while relying on humans.
CATA does not truly distinguish the use of algorithmic coder as a set of research processes

44
Q

Blending computers with a human

A

Hybrid approach. Encourage these in conducting content analysis
> improve efficacy and reduce human errors.

Distinguish studies that rely on algo orders for classification.

45
Q

Labeling distinct research methods:

A

Drawing attention to fundamental different research designs required for human versus algorithmic coders > concerns these methods raise concerning validity

46
Q

Algo coders

A

Assessed, cleaned and prepared for analysis - relevant data assessed, precision of keywords, computers can code false positives, significant error to data

47
Q

human:

A

Sacrifice narrow precision for sake of broad recall, human can pick up false positives= broadest search with greatest number of relevant articles preferred.

48
Q

ATA:

A

Stemming the text= noun, verb etc are coded to the same activity.
Function words (the, a, an etc) are removed from text before

49
Q

Advantages and dis of ATA

A

Advantage: algo coders can quickly analyse large sets of data, big data research. Traditional tools are difficult to analyse with
fast speed of coding

Algo coder is 100% reliable. validity=determined by the validity of the conceptualisation of variables by researcher and of operationalisation in the code that algorithmic coder follows.

50
Q

ALGO CODER is no more objective than human coder. Algo coder execute commands with perfect fidelity
Algo: efficacy and reliability and natural language processing and AI - improving computers to process complex human language.
Still relies on gold standard training sets coded by humans, used in machine learning

A
51
Q

Validity in algo coders can classify complex human language primary concern. Some computer methods are no better than random coin flips.

A

ATA is rarely used in studies of visual data because those media pose their own challenges for algorithms.

52
Q

ATA best applied?

A

Best for manifest variables.

53
Q

LIWC

A

Classifies word usage based on a predetermined set of dictionaries - each represents a different psychological meaning category of language.

54
Q

Dictionary approached

A

Common in sentiment analysis, classify emotional valence of text based on the frequency of the use of either positive or negative words
issues with this: give no information about how words are used in context.

KEYWORD in context approaches improve in giving words more meaning. Which words occur more frequently together? Cluster together or co-occur, may not capture the full meaning of human language.

55
Q

Natural language processing, machine learning and AI:

A

Computational linguistics and computer science disciplines are working toward programming computers that can understand human language

Even though it advances: differences in driving those who build algo tools.

56
Q

Used for data mining:

A

Discovering and extracting relationships that may exist in the data.

57
Q

Hybrid or Computer aided content analysis

A

Studies using mainly human coders to categorise but still uses computers= hybrid approach.
= computer aided text analysis - uses computers in studies in design but not fully automates coding itself.

58
Q

Computers:

A

Retrieve content from database and social media content, freeze ever-changing, dynamic web content and create a database of static observations.
Sift data (identify tweets) to organize coding tasks and coding sheets and validate those codes.

59
Q

Left-side and right-side

A

Coding sheet on right side: dropdown menus eliminate chance of invalid, values of the variables.
Cutting back on potential human biases, relying on human coders who can see full richness and complexity of human language.

60
Q

Why is twitter popular in CA?

Twitter’s API and parsing various fields in the data: straightforward. less structured data: more sophisticated programming to scrape data.

A

= structured data - each tweet various data points - stored in consistent, structured database, applied with API = users can call fields from twitter databases and store them in table format. x
questions: too many fake accounts etc

61
Q

Unstructured (individual websites), not stores in predetermined ways

Online content: semi-structured data - relatively consistent, recognisable

A

Challenge of online communication content: very little is structures,
dynamic, continuously updated

62
Q

The size and complexity of data sets are increasing. how can it be scaled up and still focus on data validity?

A

Crimson Hexagon: understand large social media data sets.

Advantage: all tweets and historical data not available for users of the open Twitter.= steep costs.

Difficult to replicate: replication is vital- against human biases

Not replicable because: twitter data proprietary, restricted service user agreement.

Hexagon is not widely available to other researchers due to high costs
algorithm to classify content is property and confidential

63
Q

Best practice requirements for the use of computer platforms (hybrid) and ATA (preserve values, replicability)

A
  • Platform scalable, big data
  • Should be a free and open source (black box)
  • Adaptable to wide range of projects (open-source)
  • Give users advance control, and also provide easy interface for novice users
  • Publicly assesses data
64
Q

Summary:

A

Computers improve efficacy, more complex datasets and reliability (validity). Human coders will be obsolete by the algo coder.
Challenge notion of human and algorithmic coders will produce equivalent data, much less identical data or fail to produce equivalent data, it is necessarily a fault of the algorithmic coder. Distinguish use of algorithmic coders as a different method by assigning a different name: ATA. best for text, rather than visual and in particular manifest variables.

65
Q

Computers enhance efficiency, and reliability of human coders.
How it can be scaled up, further developing tools making CA more efficient. Validity is the primary concern, but also other aspects like the role of peer review and replicability

A

Content analysis (11 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions