SOC200 - Conceptualization (Chapters 4+5) Flashcards
UNITS OF ANALYSIS
- what/who being studied
- Identifying can difficult
- Periodically diff betw who/what we observe to get info we need + who/what generalizing about based on info
- what characteristics interested in observing
UNITS OF ANALYSIS
•info from individuals + pooling info to find patterns
•Individuals unit of observation, but aggregates units of analysis
•Derive characteristic of social groups from those of individual members - Couples, cities, gangs, churches
Social artifacts: Concrete items – books, paintings, buildings/Social interactions – Weddings, friendship
UNITS OF ANALYSIS AND FAULTY REASONING: ECOLOGICAL FALLACY
assume observations at broader unit of analysis say something about indivs comprising unit/equally distributed to individuals in that group
Eg: Neighbourhoods with lots Protestants higher suicide rate than lots of Catholics
•couldn’t infer from aggregate level data because you don’t have individual data
•individualistic fallacy: bias, racism, prejudice
•can’t generalize characteristic of one person to whole groups
CONCEPTUALIZATION
process of agreement about what concepts mean
•Indicators & Dimensions: Sign of presence/absence of concept
•Definitions: Real, Nominal + Operational
•Conceptual Order
•If common images changes over time, then concept means something else
OPERATIONALIZATION
Specifying exact procedures involved in measuring variables represent concepts (the operational definition)
•Operationalization Choices
•Levels of Measurement
MEASUREMENT
measures use depend on operational definition + way available data collected
•Accuracy vs. Precision
•Reliability vs. Validity
•Indexes vs. Scales
CONCEPTUALIZATION: Interchangeability of Indicators
doesn’t matter if researchers disagree on specific indicators only matters if all indicators used do not behave same way in testing
CONCEPTUALIZATION: Interchangeability of Indicators
If some indicators behave differently, then possible they represent diff dimension of the concept/don’t represent that concept at all.
•Example: all indicators measure level of reckless driving, if you dropped, one you can still measure it
CONCEPTUALIZATION: Real Definitions
statement of essential nature of some entity
IMPOSSIBLE – mistakes construct for a real entity + produces vague definitions
•only exists in theory
CONCEPTUALIZATION: Nominal Definitions
arbitrary definition
based on conventions of how term commonly used everyday conceptual understanding
CONCEPTUALIZATION: Operational Definitions
specifies exactly how a concept will be measured
high clarity by limiting definition - how it is measured only means that thing
CONCEPTUALIZATION: Conceptual Order
need to refine/expand meaning, their increased understanding of the concept necessitates it
may divide up concept into diff dimensions + subdimensions
•Hermeneutic circle: cyclical process of ever deeper understanding
OPERATIONALIZATION: Range of Variation
what distance betw highest + lowest points in data
•Example from HBSC Survey: frequency distribution, maybe decide $100+
oHow much variation do you need?
OPERATIONALIZATION: Degree of Precision
degree of precision at the time of data collection will affect what you can and cannot do with the data later
you can always break it down into less precise units
•If you don’t need to know exactly how much they smoke, then put it into categories
Degree of Precision is Important
cannot collect more general data then break it down to specific later
•More precise data has more informative models
•General data limits types of models you can use
•If possible more detail, can aggreagate later
•More precise – more measurement validity
•Collapsing data is more vague, and less measurement power
LEVELS OF MEASUREMENT: Nominal Measure
Variables with attributes merely categorical, no rank, no numeric value
different from each other and have no order
Yes No
LEVELS OF MEASUREMENT: Ordinal Measure
Variables with attributes rank ordered
can’t really pin down range/distance between attributes
Never, Occasionally, Regularly, Frequently, All the time, or 0-10, 11-20, etc.
•Ordinal - percentage
LEVELS OF MEASUREMENT: Interval Measure
Variables with attributes can rank + equal distances, no true 0
IQ Score 60,70,80,90,100,110,120,130
LEVELS OF MEASUREMENT: Ratio Measure
Interval measure but with a true 0 point
income, kelvin-0,1,2,3,4,5,6,7,8,9,10,11,12,13,
•Always nice to collect data at ratio level
•Certain types of analysis requires certain level of measurement
•Conclusions have to reflect data you have
•Ratio–mean
Reliability
- Consistent = reliable
- We don’t know whether 1 time measurement affected by other factors
- Replication important
Validity
-Exists when measures accurately reflect the concept it is intended to measure
•Measure that accurately reflects concept it is intended to measure
Reliable, not valid
consistent, but don’t hit target
oConcepts require rethinking
Valid, not reliable
measures scattered, one point, not consistent in measurement
Neither valid, nor reliable
no point in target, not very well clustered
Both valid + reliable
points clustered around target, Consistent and close to concept
Methods for Creating Reliable Measures: 1.Test-Retest Method
Make the same measurement more than once
Methods for Creating Reliable Measures: Split-Half Method
Good for measuring complex concepts comprised of several indicators (good health, religiosity, prejudice).
Statistical method to see how correlated indicators are: Randomly assign half the indicators one set + other indicators as another set
If highly positively correlated, measuring same thing + are reliable measures of your concept.
Methods for Creating Reliable Measures: Using Established Measures
•Using established measures: previously used + known reliable scales
Rely instead on measures that have been proven in previous research
Methods for Creating Reliable Measures: Reliability of Research Workers
Ensuring research workers (interviewers, coders, transcribers, translators, analysts) performing as consistently as possible
trained workers as coding the same way
Methods for Creating Reliable Measures: Reliability of Research Workers
calling subsample of respondents to verify certain info
have same piece of info independently coded by more than one worker or plenty of training
•Clarity, specificity, training, and practice
GUIDES FOR VALID MEASUREMENT: Face Validity
Indicator seems sensible measure of the concept
checking if they attend class
GUIDES FOR VALID MEASUREMENT: Criterion-related (predictive)Validity
Based on an external criterion
employee hiring, sats
GUIDES FOR VALID MEASUREMENT: Content Validity
How much does measure cover range of meanings in concept?
includes passive receiving on sexual activity
GUIDES FOR VALID MEASUREMENT: Construct Validity
How does the variable relate theoretically to other variables?
created a measure using marks + associate with hours of study
•measure should relate in some logical fashion to other measures
COMPOSITE MEASURES: INDEXES
measures based on more than 1 data item
summarizing + rank scores assigned to individual indicators Usually represents a more general dimension
INDEXES: Must have face validity
item should measure what you intend it to measure
13 HBSC items all appear to offer some indication of the person’s health
INDEXES: Unidimensionality
Each item only represent 1 dimension of concept
items measuring physical health should not be included with items measuring mental health
INDEXES: Decide on Specificity of the Variable
only want to measure certain aspect of health, then only use items representing that aspect
If measuring overall health, then include more balanced set of items
INDEXES: Look at Amount of Variance each Item Provides
If your item has 100% of people answering the same way, the item will have no variance and will not be a useful indicator for your measure
COMPOSITE MEASURES: SCALES
assigns scores to patterns of responses, some items suggest weak degree of the variable + others reflect stronger degrees of the variable
Takes advantage of diff in intensity among indicators
•more variation
Scales and Indexes – Similarities
▫Both ordinal measures of variables
▫Both composite measures of variables
•Both rank order units of analysis using specific variables
Scales and Indexes – Differences
▫Index-accumulating the scores assigned to individual items
•index more cumulative
oSimply accumulating scores assigned to individual indicators
Scales and Indexes – Differences
▫Scale assigns scores to patterns of responses suggesting diff in intensity among indicators of the variable
Conceptualizing/Operationalizing “Good Health”
- Indicators: eats well, smiling, fit
- Working definition: no physical/mental conditions that proclude enjoyment of life
- Agreed ideas on what concepts mean come from everyday observation
- Observations make up conceptions
- Agreed upon conceptions make up definition
- Going from mental images (conceptions) to indicators and dimensions
Ambiguity of Meaning in a Concept
- drawing on everyday meaning of sexual relationship
- got off on lewinsky case
- paula jones – lawyers pins down with declarative (written) definion
- Clinton’s answers were legally correct
- Ambuiguity researchers can face with definitions
- Can sabotage judicial procedures
- Left a loophole for Clinton
Construct
theoretical creations based on observations but cannot be observed directly or indirectly
•Concepts are constructs derived by mutual agreement from mental images
•Concept have no real, true, or objective meanings
help us organize, communicate about, and understand things that are real
